Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12

ADVANCES IN THE ECONOMIC ANALYSIS OF PARTICIPATORY AND LABOR-MANAGED FIRMS ADVANCES IN THE ECONOMIC ANALYSIS OF PARTI...

Author: Jed Devaro

129 downloads 725 Views 1MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

ADVANCES IN THE ECONOMIC ANALYSIS OF PARTICIPATORY AND LABOR-MANAGED FIRMS

ADVANCES IN THE ECONOMIC ANALYSIS OF PARTICIPATORY AND LABOR-MANAGED FIRMS Series Editor: Takao Kato Recent Volumes: Volume 8:

Employee Participation, Firm Performance and Survival – Edited by V. Perotin & A. Robinson

Volume 9:

Participation in the Age of Globalization and Information – Edited by Panu Kalmi & Mark Klinedinst

Volume 10: Cooperative Firms in Global Markets: Incidence, Viability and Economic Performance – Edited by Sonja Novkovic & Vania Sena Volume 11: Advances in the Economic Analysis of Participatory and Labor-Managed Firms – Edited by Tor Eriksson

ADVANCES IN THE ECONOMIC ANALYSIS OF PARTICIPATORY AND LABOR-MANAGED FIRMS VOLUME 12

ADVANCES IN THE ECONOMIC ANALYSIS OF PARTICIPATORY AND LABOR-MANAGED FIRMS EDITED BY

JED DEVARO Department of Management and Department of Economics, California State University East Bay, Hayward, CA, USA

United Kingdom – North America – Japan India – Malaysia – China

Emerald Group Publishing Limited Howard House, Wagon Lane, Bingley BD16 1WA, UK First edition 2011 Copyright r 2011 Emerald Group Publishing Limited Reprints and permission service Contact: [email protected] No part of this book may be reproduced, stored in a retrieval system, transmitted in any form or by any means electronic, mechanical, photocopying, recording or otherwise without either the prior written permission of the publisher or a licence permitting restricted copying issued in the UK by The Copyright Licensing Agency and in the USA by The Copyright Clearance Center. No responsibility is accepted for the accuracy of information contained in the text, illustrations or advertisements. The opinions expressed in these chapters are not necessarily those of the Editor or the publisher. British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library ISBN: 978-0-85724-759-9 ISSN: 0885-3339 (Series)

Emerald Group Publishing Limited, Howard House, Environmental Management System has been certified by ISOQAR to ISO 14001:2004 standards Awarded in recognition of Emerald’s production department’s adherence to quality systems and processes when preparing scholarly journals for print

CONTENTS LIST OF CONTRIBUTORS

ix

FOREWORD

xi

INTRODUCTION

xiii

PART I: JOB DESIGN AND ORGANIZATIONAL PERFORMANCE SPECIALIZATION, MULTISKILLING, AND ALLOCATION OF DECISION RIGHTS Hideo Owan

3

THE EFFECT OF MULTISKILLING ON LABOR PRODUCTIVITY, PRODUCT QUALITY, AND FINANCIAL PERFORMANCE Martin Farnham and Emma Hutchinson

35

TEAMS, AUTONOMY, AND THE FINANCIAL PERFORMANCE OF FIRMS: NEW EVIDENCE FROM PANEL DATA Jed DeVaro

63

PART II: COMPENSATION, WORKER ATTITUDES, AND PRODUCTIVITY HOW DO RULES AND COSTS AFFECT A FIRM’S SETTING OF BENEFITS? THE CASE OF HEALTH INSURANCE AND WORKFORCE SKILLS Nan L. Maxwell v

89

vi

CONTENTS

MAJORITY OWNERSHIP AND CHIEF EXECUTIVE COMPENSATION Derek C. Jones and Niels Mygind

115

WORKER ATTITUDES TOWARD EMPLOYEE OWNERSHIP, PROFIT SHARING AND VARIABLE PAY Fidan Ana Kurtulus, Douglas Kruse and Joseph Blasi

143

PERFORMANCE-RELATED PAY, UNIONS, AND PRODUCTIVITY IN ITALY: EVIDENCE FROM QUANTILE REGRESSIONS Mirella Damiani and Andrea Ricci

169

PART III: WORKER COOPERATIVES AND NONPROFIT ORGANIZATIONS PROFIT REINVESTMENT IN ITALIAN WORKER COOPERATIVES AS A CONTRIBUTION TO A COMMON GOOD: AN EMPIRICAL ANALYSIS ON WORKERS’ PERCEPTION AND MOTIVATION Cecilia Navarra

199

DOES TRAINING POLICY HELP TO ATTRACT, RETAIN, AND DEVELOP VALUABLE HUMAN RESOURCES? ANALYSIS FROM THE MONDRAGON CASE Imanol Basterretxea and Eneka Albizu

231

GOVERNANCE AND BEHAVIOR IN NONPROFITS: ANALYSIS OF URUGUAYAN HEALTH CARE ORGANIZATIONS Juan Jose Barrios and Mieke Meurs

261

Contents

vii

PART IV: FREE TRADE AND THE ECOLOGICAL EFFECTS OF ALTERNATIVE SOCIO-ECONOMIC SYSTEMS CAPITALISM, ECONOMIC DEMOCRACY, AND ECOLOGICAL DESTRUCTION OF OUR PLANET Jaroslav Vanek

289

THE CASE FOR CAPITALISM: A COMMENT ON JAROSLAV VANEK’S ‘‘CAPITALISM, ECONOMIC DEMOCRACY, AND ECOLOGICAL DESTRUCTION OF OUR PLANET’’ Jed DeVaro and Adrian Stoian

299

REJOINDER Jaroslav Vanek

307

LIST OF CONTRIBUTORS Eneka Albizu

Universidad del Paı´ s Vasco, EU Relaciones Laborales, Leioa, Spain

Juan Jose Barrios

Department of Economics, ORT University, Montevideo, Uruguay

Imanol Basterretxea

Universidad del Paı´ s Vasco, Facultad de Ciencias Econo´micas y Empresariales, Bilbao, Spain

Joseph Blasi

School of Management and Labor Relations, Rutgers University, Piscataway, NJ, USA

Mirella Damiani

Department of Economics, Finance and Statistics, University of Perugia, Perugia, Italy

Jed DeVaro

Department of Management and Department of Economics, California State University East Bay, Hayward, CA, USA

Martin Farnham

Department of Economics, University of Victoria, Victoria, British Columbia, Canada

Emma Hutchinson

Department of Economics, University of Victoria, Victoria, British Columbia, Canada

Derek C. Jones

Department of Economics, Hamilton College, Clinton, NY, USA

Douglas Kruse

School of Management and Labor Relations, Rutgers University, Piscataway, NJ, USA

ix

x

LIST OF CONTRIBUTORS

Fidan Ana Kurtulus

Department of Economics, University of Massachusetts-Amherst, Amherst, MA, USA

Nan L. Maxwell

Mathematica Policy Research, Oakland, CA, USA

Mieke Meurs

Department of Economics, American University, Washington, DC, USA

Niels Mygind

Department of International Economics and Management, Copenhagen Business School, Frederiksberg, Denmark

Cecilia Navarra

Department of Economics ‘‘Cognetti de Martiis’’, University of Torino, Torino, Italy

Hideo Owan

Institute of Social Science, The University of Tokyo, Bunkyo-ku, Tokyo, Japan

Andrea Ricci

Area di Analisi e Valutazione delle Politiche per l’Occupazione, ISFOL, Rome, Italy

Adrian Stoian

Department of Economics, California State University East Bay, Hayward, CA, USA

Jaroslav Vanek

Department of Economics (Emeritus), Cornell University, Ithaca, NY, USA

FOREWORD The series Advances in the Economic Analysis of Participatory and LaborManaged Firms was launched over 25 years ago by Derek C. Jones and Jan Svejnar. Since then, Advances has been a leading forum for high-quality original theoretical and empirical research in the broad area of participatory and labormanaged organizations. Although general and specialized journals publish work in this ﬁeld, many do so only occasionally. Advances has been the only annual periodical that presents some of the best papers in the ﬁeld in a single volume. It is my great pleasure to present Volume 12 of Advances in the Economic Analysis of Participatory and Labor-Managed Firms. Advances has been making frequent use of guest editors. This volume is also ably edited by Jed DeVaro. The selection of Jed as this volume’s guest editor symbolizes the renewed recognition of the importance of theoretical work in the ﬁeld (Jed is among a few young economists who have been making signiﬁcant contributions to the ﬁeld not only empirically but also theoretically). It is my hope that Advances increasingly reﬂects such renewed interest in theoretical work which will guide empirical studies in more fruitful directions. The scope of Advances will also continue to reﬂect great changes in the realities of participatory organizations in the last two decades or so. Following the disintegration of the Former Republic of Yugoslavia, the principal systemic example of self-management was replaced with diverse forms of participatory systems. In advanced market economies, many ﬁrms have been experimenting with new and innovative work practices aimed at promoting employee participation in decision making in the workplace (sometimes even at the top corporate level) and alternative compensation systems designed to align the interest between labor and management. In addition, a number of signiﬁcant examples of worker cooperatives have ﬂourished. In transition economies, the collapse of the former USSR triggered widespread experimentation with diverse forms of participation, in particular employee ownership. I hope you will ﬁnd this volume informative and stimulating and that you will consider contributing to the future volumes of Advances and sharing information about Advances with other interested colleagues. Takao Kato Series Editor

xi

INTRODUCTION Continuing in the tradition of earlier volumes in the series, the 12th edition of Advances in the Economic Analysis of Participatory and Labor-Managed Firms interprets the themes of the series quite broadly and, consequently, the following pages traverse a highly variegated terrain. I have marked this expansive landscape with four guideposts around which our discussion is organized, namely sections on job design and organizational performance (Part I); compensation, worker attitudes, and productivity (Part II); worker cooperatives and nonproﬁt organizations (Part III); and free trade and the ecological effects of alternative socio-economic systems (Part IV). The following essays exhibit a diversity of topics, data sources, modes of analysis, geographic contexts, and philosophies. The contributors are similarly diverse, hailing from seven countries, with representation both inside and outside of academia. It is my hope that, in addition to contributing to our knowledge of the broad subject at hand, the articles contained in this volume will inspire productive future work on the important questions addressed herein. The leading article of the volume, and of Part I – which addresses issues of job design and organizational performance – is the jewel by Hideo Owan entitled ‘‘Specialization, multiskilling, and allocation of decision rights.’’ By developing a model that considers the effect of multiskilling on the levels of noncontractible human capital investment, delegation of authority, and unionization, Owan offers another rationale for the multiskilling practices that have been increasingly observed in ﬁrms in recent decades. Owan’s key insight is that worker investments in ﬁrm-speciﬁc human capital become strategic substitutes when their skills overlap each other. Workers compete for bargaining power more intensively when their skills overlap than when they do not. Because of the ‘‘competition effect’’ created by this substitutability, unless specialization offers a substantial technological advantage, then with multiskilling (as opposed to specialization of labor) the following is true: (1) workers’ incentives to invest in ﬁrm-speciﬁc human capital tend to be stronger; (2) the optimal level of delegation of authority is typically higher; (3) ﬁrms’ ex post proﬁts tend to be higher. Firms typically delegate authority to an inefﬁciently low extent, due to the fear of relinquishing too much bargaining power to workers, but Owan’s analysis shows that multiskilling mitigates this distortion. xiii

xiv

INTRODUCTION

Investigating the practice of multiskilling through an empirical rather than a theoretical lens takes us to the second chapter, where Martin Farnham and Emma Hutchinson guide us through an analysis of ‘‘The effect of multiskilling on labor productivity, product quality, and ﬁnancial performance.’’ Using data from the 2004 British WERS, a nationally representative cross-section of establishments, they measure the effect of multiskilling on establishment-level labor productivity, product quality, and ﬁnancial performance, ﬁnding that accounting for the endogeneity of multiskilling in the empirical models (via bivariate probit analyses) has signiﬁcant implications for the results. In particular, the estimated (positive) effect of multiskilling on labor productivity disappears after accounting for the endogeneity of multiskilling. Moreover, the estimated effect of multiskilling on product quality goes from zero to positive, and the (positive) effect on ﬁnancial performance increases signiﬁcantly in magnitude. The results are important, both because empirical work on the effects of multiskilling on organizational performance is relatively scarce and because such work rarely if ever accounts for endogeneity of multiskilling. The job design discussion then turns from multiskilling to teams, with my work in the third chapter entitled ‘‘Teams, autonomy, and the ﬁnancial performance of ﬁrms: New evidence from panel data.’’ The goal of this chapter is to re-examine, using panel data, the topic I studied ﬁve years ago using cross-sectional data, in a paper entitled ‘‘Teams, autonomy, and the ﬁnancial performance of ﬁrms.’’ In the present chapter, I use the 1998–2004 British WERS data to analyze the effect of team production (either autonomous/self-managed or closely managed) on ﬁnancial performance. The pattern of evidence is consistent with a positive association between team production in an establishment’s largest occupational group and the likelihood of improved ﬁnancial performance for that establishment. However, the results are mixed concerning whether the positive effects of teams are larger for autonomous teams versus nonautonomous teams. Part II of the volume addresses issues of compensation, worker attitudes, and productivity. A key ingredient in high-involvement and high-performance work systems is premium compensation, which includes generous fringe beneﬁts. A central idea in the economics literature on fringe beneﬁts is the notion of a wage-beneﬁts tradeoff, which is predicted by the theory of compensating differentials. The point is that, in equilibrium, ﬁrms and jobs in which relatively generous fringe beneﬁts are paid should be characterized by correspondingly lower wages. From an empirical standpoint, the main challenge to testing for the existence of such substitution is the difﬁculty of controlling sufﬁciently for relevant worker, ﬁrm, and job characteristics.

Introduction

xv

Even though a number of studies have made impressive attempts to control for such factors comprehensively, an empirical tradeoff between wages and beneﬁts has remained elusive. This naturally leads one to wonder whether there is more to the story than unobserved heterogeneity in workers and employers. Perhaps some other force is contributing to obfuscation of the empirical tradeoff. A natural place to begin looking for such a culprit is by exploring the relevant institutions. Nan Maxwell does just that, in her chapter entitled ‘‘How do rules and costs affect a ﬁrm’s setting of beneﬁts? The case of health insurance and workforce skills.’’ Maxwell examines how institutional rules and economies of scale can create incentives for ﬁrms to make inframarginal decisions when offering health insurance. Using unique data from the California Health and Employment Survey, she ﬁnds results suggesting that failure to control for such inframarginal employer decisions contributes to a failure to empirically identify the wage-insurance tradeoff predicted by ‘‘institution free’’ theory. Among other results, Maxwell documents a 10–13 percentage point increase in the probability of an establishment offering workers health insurance in jobs outside of those in which compensation is being set, if the recruiting difﬁculty lies in mid- or high-skilled positions. This increase is about twice the size of the increase associated with recruiting difﬁculty in the position in which compensation is negotiated. One of the lessons from this work, which will hopefully inﬂuence the direction of future inquiry on the subject, is that efforts to measure the empirical tradeoff should focus on the employer level rather than the individual worker level. Given the prevalence of powerful institutions in the arena of employer-provided health insurance, it seems almost inevitable that they would heavily inﬂuence the wage-beneﬁts tradeoff. More generally, the question of how employers respond to institutional incentives when designing compensation systems is of critical policy importance, particularly given the heated present debate in the United States concerning how health insurance should be provided. Although the workers in Maxwell’s data are nonexecutives, scaling the rungs of the organizational job hierarchy to its peak takes us to the next chapter. There is a large empirical literature on the topic of executive compensation, most of which focuses on Western countries, particularly the United States. Very few studies have addressed compensation in the former Soviet economies and, until the publication of this volume, no work had been done on executive pay practices in Estonia. Jones and Mygind produce the ﬁrst such work in their chapter entitled ‘‘Majority ownership and chief executive compensation.’’ The authors exploit unique panel data from Estonia to explore the effects of differing types of majority ownership,

xvi

INTRODUCTION

including employee ownership, on executive compensation. One of their main results is that an economically signiﬁcant determinant of executive compensation is ownership, in both state and privatized ﬁrms as well as among different types of private ﬁrms. A second result is that in ﬁrms with majority ownership by employees, compensation is about 15 percent less than in state-owned ﬁrms, ceteris paribus. A third result speaks to the oftexplored relationship between ﬁrm size and executive compensation. In particular, the authors ﬁnd much smaller ‘‘ﬁrm size elasticities’’ than those estimated in the previous literature based mostly on ﬁrms in Western countries. Overall, their results both highlight the varying importance of principal–agent relationships across different types of ﬁrms and support a conclusion that privatization has imposed strong discipline on the level of executive compensation. Returning to nonexecutives, the preferences of such workers over various forms of compensation are a central determinant of how employers structure their compensation offerings. An exploration of such worker preferences is undertaken by Fidan Ana Kurtulus, Joseph Blasi, and Douglas Kruse, in their chapter entitled ‘‘Worker attitudes toward different forms of employee ownership and variable pay.’’ The authors use the NBER Shared Capitalism Database, comprised of over 40,000 employee surveys from 14 ﬁrms, to investigate worker preferences for employee ownership, proﬁt sharing, and variable pay using detailed survey questions pertaining to those forms of compensation. Of primary interest is how preferences for those various types of output-contingent pay vary with worker risk aversion, residual control, and perceptions of co-workers and management. The main results of the paper reveal that, on average, workers prefer to have at least part of their compensation be performance-related, with stronger preferences for output-contingent pay among workers who have lower levels of risk aversion, greater control over the work process, and greater trust of co-workers and management. The Kurtulus–Blasi–Kruse chapter focuses on worker preferences which, as noted, are a key determinant of various forms of output-based pay. In addition to considering the determinants of output-based pay, an important area of research in the empirical incentives literature concerns the consequences of such output-based pay, and in particular the consequences for performance. That topic is addressed in the context of the Italian economy in the chapter by Mirella Damiani and Andrea Ricci entitled ‘‘Performancerelated pay, unions and productivity in Italy: Evidence from quantile regressions.’’ The authors explore the relationship between performancerelated pay and productivity using a nationally representative sample of manufacturing and service companies. In the ﬁrst of two steps, the authors

Introduction

xvii

estimate a classical production function using longitudinal data on the balance sheet variables of ﬁrms during 2002–2005. In the second step, they regress the distribution of the ﬁrm-speciﬁc ﬁxed effects on indicators for performance-related pay and unions and on year-2005 controls. The main result is that adoption of performance-related pay is positively associated with productivity throughout the whole distribution and that the presence of unions is positively associated with ﬁrms’ unobserved productivity across all quantiles, being signiﬁcantly higher for those ﬁrms at the highest quantile of the productivity distribution. Part III of the volume addresses worker cooperatives and nonproﬁt organizations.In her chapter entitled, ‘‘Proﬁt reinvestment in Italian worker cooperatives as a contribution to a common good: an empirical analysis on workers’ perception and motivation,’’ Cecilia Navarra applies tools from the collective action literature to explain the high propensity of Italian worker cooperatives to reinvest proﬁts into asset locks, i.e. common funds that are nondivisible and nonappropriable by individual members, neither when they retire nor at the end of the cooperative’s life. Funds contributed to asset locks cannot be distributed to members and must be used only for the ﬁrm’s overall purposes. Navarra characterizes asset locks as a common pool resource, creating a collective action problem that may appear in two forms: free-riding in effort provision and an inter-temporal incentive problem whereby current generations have an incentive to underinvest in the common pool resource given that they will not reap the future dividends from those investments. Navarra asks what individual characteristics (i.e. collective or groupbased motivation, and attitudes toward time preference) facilitate a cooperative willingness to reinvest proﬁts in the ﬁrm. To shed light on this, she uses a survey of a sample of workers in cooperatives afﬁliated with the Legadelle Cooperative e Mutue of the province of Ravenna, Italy, to illuminate the attitudes of workers towards asset locks. She ﬁnds some evidence suggesting a higher degree of loyalty to ﬁrms that amass a greater share of proﬁts into asset locks. However, she cannot determine whether this arises because of repeated interactions or because of the employment insurance provided by the cooperative. Workers identiﬁed as having a positive attitude toward asset locks exhibit a preference for the relational aspects of the job and for ‘‘on-the-job’’ consumption. Members who participate in the decision on proﬁt allocation are found to be more willing to reinvest proﬁt into asset locks. A striking result is that employees are, on average, more likely to favour asset locks than are worker-members, which Navarra explains by noting that only members participate in proﬁt sharing.

xviii

INTRODUCTION

In his essay at the end of this volume, Jaroslav Vanek cites as an illustrative example the well-known Mondragon cooperatives of Spain. These coops are the focus of a case study on training, recruitment, and retention, conducted by Imanol Basterretxea and Eneka Albizu in their chapter entitled ‘‘Does training policy help to attract, retain, and develop valuable human resources? Analysis from the Mondragon case.’’ The authors aim to investigate the extent to which a training policy developed through corporate training centers is recognized by HR managers as a source of competitive advantage for attracting, developing, and retaining valuable employees. Their descriptive analysis is based on the case of the Spanish Mondragon cooperative group and, in particular, a survey of HR managers from 66 cooperatives in the group. The main result of the descriptive analysis is that, in the eyes of HR managers, Mondragon’s training policy is perceived to be a useful tool for attracting, developing, and retaining valuable human resources. The results also suggest that the aforementioned advantages of training policies are more modest than has been recognized in the literature on Mondragon. Focusing on the perceptions of HR managers is potentially important, to the extent that such managers inﬂuence policy regarding the recruitment, screening, and training of new workers, as well as strategies to retain new hires that reveal themselves to be of high value. It can be helpful, if we want to understand how various practices affect organizational performance, to learn the perspectives of those who administer and in some cases propose and design these practices. Data arising from survey questions of the type analyzed here potentially suffer from a variety of biases and this, along with the fact that the tabulations the authors report are basic and unconditional, should be kept in mind when interpreting the results. Among other concerns, one might worry that a manager’s perception of the success of a program is likely colored by the fact that the program was adopted in the ﬁrst place. It can be expected that HR managers may think more positively about programs that have been purposefully adopted, perhaps with their input. Concerns also arise about how managers interpret various questions, and about the scarcity of quantiﬁable effects (e.g., we might learn that HR managers believe that training improves retention, but of ultimate interest is the magnitude by which a given expenditure on training improves retention). Another concern is the survey’s undue focus on beneﬁts and not on costs. So we might learn that HR managers believe that ‘‘Offering more continuous training than competitors helps us to retain valuable staff,’’ but doubling everyone’s compensation would accomplish the same thing, and in the absence of further information, there is really no reason to

Introduction

xix

prefer one of these HR practices over the other. These data issues notwithstanding, the chapter offers food for thought and should encourage further exploration of the topic in future work. Bouncing from the Mondragon cooperatives of Spain down to the southern hemisphere takes us to the next chapter. A growing literature studies differences between for-proﬁt ﬁrms and nonproﬁt organizations. Less attention has been devoted to comparing different types of nonproﬁts or other nontraditional ﬁrms, and that is the goal of Juan Jose Barrios and Mieke Meurs in their chapter entitled ‘‘Governance and behavior in nonproﬁts: Analysis of Uruguayan health care organizations.’’ The authors use a unique data set to examine differences across three types of Uruguayan nonproﬁt health care organizations. Their main result is that the structure of stakeholding and governance signiﬁcantly affects behavior even where many behaviors are highly regulated, and their ﬁndings highlight the importance of carefully specifying stakeholder and governance structure when predicting the behavior of nonproﬁts. Part IV concludes the volume with a debate on the subject of comparative economic systems and the implications for the environment. In a provocative essay entitled ‘‘Capitalism, economic democracy, and ecological destruction of our planet’’ the venerable Jaroslav Vanek takes the stage to ignite a lively debate on the virtues and vices of capitalist economic systems versus fully democratic systems from the standpoint of the ecological wellbeing of the planet. Vanek argues that fully democratic systems of economic organization, which rely on democratic decisions based on personal rights in areas such as enterprise management and organization, are in many cases particularly well suited for protecting the natural environment. The logic is that the decision makers in such democratic systems live in and are permanently exposed to the environment, creating natural incentives to protect it. In contrast, in capitalist economic systems, ﬁrms are interested solely in maximizing proﬁts, focusing on their private costs of production as opposed to social costs and ignoring the negative environmental production externalities associated with their activities. Consequently, Vanek concludes that public regulation is of greater necessity for capitalist ﬁrms than for ﬁrms in fully democratic economic systems. In counterpoint to these arguments, I join Adrian Stoian in presenting a case for the capitalist system. In a brief rejoinder following our comment, Vanek closes the debate and also this volume with some views on the destructive potential of free trade. Jed DeVaro Guest Editor

PART I JOB DESIGN AND ORGANIZATIONAL PERFORMANCE

SPECIALIZATION, MULTISKILLING, AND ALLOCATION OF DECISION RIGHTS Hideo Owan ABSTRACT The purpose of this chapter is to offer new justiﬁcation for multiskilling practices such as job rotation and extensive training for broad skills and explain why there appear to exist complementarity between multiskilling and the delegation of decision authority to workers. By developing a new model of incomplete contracting where workers make noncontractable investments in multiple skills, we obtain the key insight that worker investments in ﬁrm-speciﬁc human capital become strategic substitutes when their skills overlap each other. The ‘‘skill substitution effect’’ analyzed in this chapter induces the following three major results, unless specialization offers a substantial technological advantage: (1) workers’ incentives to invest in ﬁrm-speciﬁc human capital tend to be stronger; (2) the optimal level of delegation is typically higher; and (3) ﬁrms’ ex post proﬁts tend to be higher with multiskilling than with specialization.

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 3–34 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012005

3

4

HIDEO OWAN

The novel implication of the chapter is that multiskilling may be desirable from a ﬁrm’s viewpoint even if there are no technological or informational task complementarities among the combined skills, which have been believed to be primary reasons for multiskilling in prior works. Keywords: Multiskilling; task complementarities; delegation; job design; wage bargaining; human capital investment JEL classiﬁcations: J24; J31; M53; M54

INTRODUCTION One of the most drastic changes in economists’ views on job design in the past few decades is their recognition of the gains from multiskilling vis-a-vis specialization.1 The idea that specialization and division of labor allow more output from a given set of inputs has been embraced by economists because numerous anecdotes (including Adam Smith’s famous description of a pin factory) and historical productivity data supported the proposition. Obviously, a number of economists, including Adam Smith, have discussed the factors that ultimately limit specialization (e.g., size of markets and complementarity among tasks).2 However, scrutinizing trade-offs affecting the optimal division of labor have become more important recently due to a current trend toward job enlargement. A number of authors have attempted to explain basic trade-offs determining job design. Becker and Murphy (1992) assert that the degree of specialization is constrained by various costs of coordinating specialized workers who perform complementary tasks and the amount of general knowledge available for each task. A similar view is presented by Bolton and Dewatripont (1994), who argue that the beneﬁts of greater specialization in information processing by having more agents team up within the same organizations are partly (and sometimes entirely) offset by the increased costs of communication within the enlarged group of agents. These theories, however, cannot explain why we are observing more workplaces adopting multiskilling practices recently. For example, they imply that falling costs of coordination and communication due to advancement in information technology should encourage specialization, not multiskilling. Koike (1977, 1988) is perhaps the ﬁrst author to emphasize the improvement of problem-solving skills as the most important beneﬁt of

Specialization, Multiskilling, and Allocation of Decision Rights

5

multiskilling. According to his extensive research of Japanese automobile plants, job rotation and other multiskilling practices prevalent in those plants help workers to understand the whole process and acquire capabilities to respond to productivity and quality issues as they arise in the workplace.3 Lindbeck and Snower (2000) also focus on a difference in learning patterns between ‘‘tayloristic’’ (characterized by specialization) and ‘‘holistic’’ (characterized by multiskilling practices) work organizations. They distinguish two types of learning, ‘‘intratask’’ and ‘‘intertask’’ learning, where the former is traditional learning-by-doing best attained by performing a narrow task, and the latter arises when a worker can use the information and skills acquired at one task to improve his performance at other tasks. Employers then face a trade-off between returns to specialization, which enhances intratask learning, and returns from task complementarities, which accumulate through intertask learning. They argue that a shift toward holistic organizations has been driven by four forces, including advances in production technologies promoting technological task complementarities, advances in information technologies promoting informational task complementarities, changes in worker preferences in favor of versatile work, and advances in human capital that make workers more versatile. Some beneﬁts of multitasking are also mentioned as part of the discussions on complementarities supporting team-based high-performance work systems. Firms adopting a certain bundle of human resource and work practices typically including autonomous teams, contingent compensation, job rotation, extensive training, and worker involvement for quality improvement are found to have experienced substantial improvement in productivity as well as product and service quality.4 Common features seen in these workplaces are an emphasis on the development of multiskilled workers, the sharing of information, and the delegation of responsibilities to teams. Milgrom and Roberts (1990, 1995) offer a coherent theory of complementarities among the ﬁrm’s choices of technologies and practices, using the framework of monotone comparative statics. They show that a fall in the costs of ﬂexible manufacturing equipment or of computer-aided design equipment lead to systematic responses including more frequent product redesigns and improvements, speedier delivery, more efforts to reduce setup and changeover costs, and so on. These changes in the ﬁrm’s incentives are expected to encourage investment in process innovations and to increase the returns to cross-training and greater autonomy for workers because the latter two practices are assumed to lower the cost of process innovations. The relative optimality of teamwork or tasksharing has also been extensively analyzed by Holmstrom and Milgrom (1991) and Itoh (1991, 1992, 1994).

6

HIDEO OWAN

In this chapter, we offer another justiﬁcation for multiskilling practices, which is distinct from the existing literature in two aspects. First, according to our theory, skill complementarities are not necessary for multiskilling to arise as the optimal practice for ﬁrms. In other words, multiskilling may be desirable from a ﬁrm’s viewpoint even if there are no technological or informational task complementarities among the combined skills. Second, our framework can explain why there appear to exist complementarity between multiskilling and the delegation of decision authority to workers, two essential elements in the high-performance work systems as discussed earlier. We use the Grossman–Hart–Moore framework of incomplete contracting where neither job design consisting of tasks, decision authority, and skill sets nor human capital investment are contractible and wages are determined by ex post bargaining. The solution concept we use is the Shapley value. Unlike Grossman and Hart (1986) and Hart and Moore (1990), the ownership structure is not our concern here. We show that an organization with a job design that relies on multiskilled workers (henceforth, an M-organization) signiﬁcantly weakens workers’ bargaining power but somewhat paradoxically raises their incentive to acquire ﬁrm-speciﬁc skills. The key insight provided by the model is that worker investments in ﬁrm-speciﬁc human capital become strategic substitutes when their skills overlap each other as a result of multiskilling. With this substitutability, it is intuitive that workers’ bargaining power relative to the owner-manager is weaker in the M-organization than in one with a specialized work force (henceforth, an S-organization), other things being equal. Less intuitive is that workers are more motivated to invest in ﬁrm-speciﬁc skills in the M-organization. This incentive comes from the fact that a worker’s broad skills in the M-organization become more valuable when other workers whose skills overlap with the ﬁrst one are absent. Note that the Shapley value of worker i, our solution concept of bargaining outcome, can be expressed as the weighted average of marginal contribution of worker i in all possible subgroups. With skill substitutability, a worker’s broad skills in the M-organization tend to be more valuable in small subgroups containing the owner of the ﬁrm than in the S-organization. As a result, a worker’s wage expressed as the Shapley value tend to be more sensitive to the accumulation of ﬁrm-speciﬁc human capital in the M-organization than in the S-organization. To sum up this argument, human capital substitutability in the M-organization lowers the return to bargaining, but raises the elasticity of bargaining returns with respect to human capital investment. Intuitively, workers compete for bargaining power more intensely when their skills overlap than when they do not.5,6

Specialization, Multiskilling, and Allocation of Decision Rights

7

One question that needs to be addressed concerning the setup of our theory is whether skill overlapping is in fact what we observe in those organizations that have adopted multiskilling practices. Given the ﬁxed numbers of workers and skills needed, multiskilling seems to require a greater degree of overlapping. One notable example that demonstrates the prevalence of increasingly overlapping skills is autonomous work teams that are perceived as an essential element of the ﬂexible work systems discussed earlier. In a typical autonomous work team, workers make decisions together and coordinate activities, and to facilitate such a high level of coordination, team members need to know one another’s job very well and share information. Job rotation is used to make members learn one another’s job. Here, overlapping skills are not just a consequence of multiskilling. Quite the contrary, the latter is introduced to achieve the former. In our theory, whether overlapping skills enhance productivity or not does not matter, but they do generate a ‘‘skill substitution effect,’’ which drives all of our results. We believe that many accounts of ﬂexible work systems (see note 4 for references) sufﬁce as evidence of a prevailing tendency toward skill overlapping. Although the mechanism we ﬁnd is innovative as an explanation for the rationale for multiskilling practices, a similar mechanism has been identiﬁed in other contexts. For example, Edlin and Hermalin (2000) analyze the situation where both the principal and the agent invest in an asset that has greater value if owned by the principal than by the agent. They show that when the principal and agent’s efforts are substitutes, the possibility of renegotiation encourages the agent to overinvest to strengthen his bargaining position. This research is also closely related to that of Stole and Zwiebel (1996b), which studied how organizational design and technology choice by a ﬁrm with no binding employment contracts differ from those of a neoclassical ﬁrm with no ex post wage bargaining. Our work enriches their model by endogenizing investment in ﬁrm-speciﬁc human capital and examines how organizational design affects the workers’ skill formation in a ﬁrm with no binding employment contracts. We further extend our framework to analyze managers’ decisions on how much authority they should delegate to the workers. Aghion and Tirole (1997) have developed a theory of the allocation of decision rights (formal authority) in situations where the decision-maker, if uninformed, can communicate with the informed party, who in such a case has effective control over decisions (real authority). In their model, the principal can provide more incentives to the agent by delegating formal authority to him (at the cost of losing control) because it raises the agent’s chance of gaining

8

HIDEO OWAN

effective real control and thus more private beneﬁt. Similarly, in our model, the delegation of more decision authority motivates workers to invest more because capabilities to make good decisions become more valuable if the workers have more decision rights. Our ‘‘incomplete contract’’ assumption induces our key insight that the owner-manager typically delegates a less than optimal amount of responsibilities to the workers. Results similar to this have been obtained by Freeman and Lazear (1995) and Prendergast (1995). Our novel ﬁnding is that the substitutability of worker skills in the M-organization encourages the manager to delegate more decision rights because the marginal wage increase due to delegation is smaller but an increase in the workers’ incentives to acquire skills is greater in the M-organization. For a large set of production technology, the optimal level of delegation is typically higher in the M-organization. As our basic model assumes individual bargaining and individual decision-making on skill investment, it is natural to ask how collective bargaining and jointly choosing skill investment might affect our results. As Horn and Wolinsky (1988) and Stole and Zwiebel (1996b) have already proved, workers gain by forming a union and engaging in collective bargaining when their human capital stocks are substitutable. Therefore, the workers in the M-organization can expropriate more surplus by unionizing ex post. However, unionization harms the workers’ incentive to acquire skills because union wage as deﬁned as an outcome of collective bargaining is less sensitive to individual human capital investment. Worker cooperation in choosing skill investment should be encouraged in a unionized ﬁrm because it mitigates free-riding, while it should be discouraged in a nonunionized ﬁrm because cooperation eliminates the ‘‘competition’’ for bargaining power in the skill investment stage discussed earlier. This implies that workers should not be involved in planning training and should be rotated between teams to discourage cooperation in skill investment in a nonunionized setting, but the converse is true in a unionized ﬁrm. We begin the second section by introducing the simple model in which two workers solve two different decision problems that require distinct skill sets. Workers make noncontractable efforts to acquire skill sets that enhance the probability of ﬁnding the best solutions for respective decision problems. We show that both specialization and multiskilling could appear in the equilibrium. In the third section, we provide the basic ﬁndings of examining the effect of multiskilling on workers’ wages and the levels of their skill investment. In the fourth section, we extend the model so that the ownermanager of the ﬁrm allocates the set of decision rights between herself and the

Specialization, Multiskilling, and Allocation of Decision Rights

9

workers to see the effect of multiskilling on the optimal allocation of decision authorities. In the ﬁfth section, we consider the effect of collective bargaining and worker cooperation in choosing skill investment on the equilibrium level of human capital, which offer a number of implications about unionization, worker training, and job rotation. The sixth section provides a general discussion of what factors affect the ﬁrm’s choice of organizational form and the seventh section offers concluding remarks.

BASIC MODEL This study deals with two distinct policies of job design: one under which workers are specialized in one task and encouraged to acquire deep, narrow skills; and one under which workers are multitasking and motivated to learn broad skills that overlap with others. We call the former organization the Sorganization and the latter the M-organization. Unlike Grossman and Hart (1986) and Hart and Moore (1990), ownership structure is not our concern. Hence, we simply assume that an owner-manager owns assets for production and hires two homogeneous workers in the market.7 The ownermanager and the workers are denoted by M, and i (i ¼ 1,2), respectively. Both the manager and the workers are risk-neutral.

Technology Suppose the ﬁrm needs to solve two decision problems denoted by k where k ¼ 1,2. The outcome of each decision is either success or failure and the ﬁrm’s output from a successful decision is m while a wrong decision generates no output. Each decision problem k requires a different ﬁrmspeciﬁc and problem-speciﬁc skill set Ak. A worker who invested in Ak can solve problem k better. Let xi,k denote worker i’s investment in skill set Ak, and let xi ¼ (xi,1, xi,2) and X ¼ (x1, x2). The personal cost of investment is c(xi). Worker i with investment xi,k can ﬁnd a right solution for problem k with probability P(xi,k). P is concave and twice continuously differentiable, and P(0) ¼ 0. We assume that the discovery of a solution is stochastically independent across workers. In other words, workers obtain solutions based on their idiosyncratic private information observed in the workplace. Let v denote the value the owner-manager can create using her assets and workers with no skills (i.e., xi,k ¼ 0 for all i and k.). We can set v ¼ 0 for the moment but later assume that v depends on the allocation of decision rights.

10

HIDEO OWAN

The probability that either worker ﬁnds a right solution for problem k is 1(1P(x1,k))(1P(x2,k)). Then, the expected joint output is as follows: YðXÞ ¼ v þ m

2 X

½1 ð1 Pðx1;k ÞÞð1 Pðx2;k ÞÞ

(1)

k¼1

The manager owns the ﬁrm in the sense of Hart and Moore (1990), and thus, the workers cannot produce anything without management. A worker with no ﬁrm-speciﬁc human capital produces nothing (i.e., P(0) ¼ 0), and the worker’s outside option value is assumed to be zero. When a worker leaves the ﬁrm after investment in ﬁrm-speciﬁc skills, he will be replaced by a new hire who has no ﬁrm-speciﬁc skills. The expected output when one or two workers leave can be easy derived. With only worker i remaining, the output is v þ m(P(xi,1) þ P(xi,2)) (i.e., xiu ¼ (0, 0) for iu 6¼ i). With both workers replaced, the surplus is v. To rule out asymmetric Nash equilibria, we impose the following assumptions: Assumption 1:

P0 ðxÞ is decreasing in x 3 2PðxÞ

Assumption 2:

cðxi;1 ; xi;2 Þ ¼ cðxi;1 þ xi;2 Þ

Assumption 1 requires that P(x) does not converge to 1 too quickly as x increases. In other words, it ensures that increased investment by one worker does not make the returns to the other’s investment fall too quickly. When Assumption 1 is violated, it could be efﬁcient to assign both decision problems to one worker and let him alone invest substantially in skills. But this case is not of much interest in light of our motivation because in such asymmetric equilibrium, there is no trade-off in specialization and multiskilling. It should also be noted that Assumption 1 is sufﬁcient to rule out any equilibria other than S-organization and M-organization as Proposition 1 notes.8 Assumption 2 assumes perfect cost substitution. By imposing Assumption 2, we can simplify our comparison of specialization and multiskilling in determining bargaining power and human capital investment. Another possibility is the cost function with skill complementarity under which acquiring one skill set helps workers to learn another skill set. Skill complementarity, however, substantially favors multiskilling, obscuring the role of human capital substitutability in multiskilling organizations. Additionally, we assume that c is convex and twice continuously differentiable and that cv(x)/cu(x) is nonincreasing with c(0) ¼ 0 and

Specialization, Multiskilling, and Allocation of Decision Rights

11

limx-Nc(x) ¼ þN. To ensure the equilibrium with positive investment, we assume the following: Assumption 3:

m 0 P ð0Þ4c0 ð0Þ 2

One example of P satisfying Assumption 3 is P(x) ¼ 1 ex, which will be frequently used throughout the chapter.

Wage Bargaining We assume that X is observable, but that neither X nor the actual outputs are veriﬁable. Hence, the manager cannot offer a contract contingent on any of these variables including the ‘‘sell the job’’ contract under which the workers pay a ﬁxed amount to the owner-manager in exchange for the income stream from their decisions. The timing of decision-making is as follows. The workers choose the level of investment x1 and x2. After the proﬁle of investment X is observed by all, the owner-manager and the workers negotiate wages. Finally, after wages are determined, the owner-manager and the workers make decisions and generate surplus. We assume that the wages are determined by the Shapley value. The Shapley value for worker i is the weighted average of his marginal contribution in all possible subgroups. The weights are calculated based on the number of all possible sequence combinations that form the subgroup.9 A non-cooperative game theoretical justiﬁcation of the Shapley value is provided by Stole and Zwiebel (1996a). In their model, the manager and the workers bargain pairwise sequentially and, in each session, play the alternating-offer bargaining game of Binmore, Rubinstein, and Wolinsky (1986), in which there is an exogenous probability of breakdown. The solution for this game is the Shapley value for the corresponding cooperative game in the limit. Later, we will also consider collective bargaining in which the union bargains with the owner-manager over the total wage. Under the bargaining structure explained above, worker i’s wage is determined by the following: m m m wi ðXÞ ¼ Pðxi;1 Þð1 Pðxj;1 ÞÞ þ Pðxi;2 Þð1 Pðxj;2 ÞÞ þ ðPðxi;1 Þ þ Pðxi;2 ÞÞ 3 3 6 (2)

12

HIDEO OWAN

Note that because the workers’ human capital investment is ﬁrm-speciﬁc, the subgroup without the owner (i.e., workers who leave the ﬁrm and form a new group) cannot produce any outputs. As a result, the bargaining return to his investment is less than the total return to the investment. This is the famous holdup problem. The proﬁt the owner-manager earns is as follows: pðXÞ ¼ YðXÞ w1 ðXÞ w2 ðXÞ m ¼ v þ ðPðx1;1 Þ þ Pðx1;2 Þ þ Pðx2;1 Þ þ Pðx2;2 ÞÞ 2 m ðPðx1;1 ÞPðx2;1 Þ þ Pðx1;2 ÞPðx2;2 ÞÞ 3

ð3Þ

Nash Equilibrium in the Investment Game First, we show that both specialization and multiskilling could arise endogenously when workers optimally choose investment in skill sets. In the equilibrium, the following ﬁrst-order conditions have to be satisﬁed: @wi m m ¼ P0 ðxi;1 Þð1 Pðxj;1 ÞÞ þ P0 ðxi;1 Þ c0 ðxi;1 þ xi;2 Þ @xi;1 3 6 @wi m m ¼ P0 ðxi;2 Þð1 Pðxj;2 ÞÞ þ P0 ðxi;2 Þ c0 ðxi;1 þ xi;2 Þ @xi;2 3 6

(4)

where i ¼ 1, 2. The following proposition proves that there could appear only two different Nash equilibria under Assumption 1: symmetric specialization and symmetric multiskilling. Proposition 1. There could be at most two Nash equilibria in the stage game where two workers choose their investment proﬁles. In one equilibrium x11 ¼ x12 ¼ x21 ¼ x22 . When m6 P0 ð0Þð3 2Pðx ÞÞ c0 ðx Þ for the solution x to the equation m6 P0 ðx Þð3 2Pð0ÞÞ ¼ c0 ðx Þ, there is another equilibrium, in which xi1 ¼ xj2 ¼ x and xi2 ¼ xj1 ¼ 0. Proof. See the appendix. Proposition 1 ensures that workers choose the same investment level and invest either uniformly between the skill sets or specialize in one skill set. There is no asymmetric equilibrium where workers choose different

Specialization, Multiskilling, and Allocation of Decision Rights

13

investment levels or invest in both skill sets but do so asymmetrically. Although ‘‘multiskilling’’ equilibrium always exists, ‘‘specialization’’ equilibrium may not exist unless there is some gain from specialization. k To make this point clear, consider PðxÞ ¼ 1 ex where k is a parameter of gains from specialization and kW0. The higher the value of k, the more likely is specialization to be efﬁcient. We can easily show that ‘‘specialization’’ equilibrium exists if kW1 but does not if kr1. Firm’s Choice of Job Design Proposition 1 allows us to focus on the two types of job design: specialization and multiskilling. In the rest of our analyses, we assume that the owner-manager can ﬁrst choose either type of job design. Such assumption is reasonable for two reasons. First, when both equilibria in Proposition 1 exist (i.e., there are some gains from specialization), the owner-manager should be able to induce the equilibrium with more proﬁtable outcome to realize by guiding workers’ beliefs. For example, management can express its expectation and offer training opportunities to encourage a certain learning behavior. Second, the owner-manager can directly control the skill sets that workers acquire by offering speciﬁc training programs or complementary practices (e.g., team activities, job rotation, and evaluation) designed for either specialization or multiskilling. We call organizations with specialized skill investment S-organization and those with multiskilling practices M-organization. Let xi be the total investment made by worker i. Namely, xii ¼ xi in the S-organization and xi1 ¼ xi2 ¼ x2i in the M-organization. Then, by substituting these investment choices into Eq. (1), the expected outputs in the S-organization and M-organization, YS and YM respectively, are expressed as follows: Y S ðx1 ; x2 Þ ¼ v þ mðPðx1 Þ þ Pðx2 ÞÞ

(5)

h x x i 1 2 Y M ðx1 ; x2 Þ ¼ v þ 2m 1 1 P 1P 2 2

(6)

Likewise, we can compare the wage functions in S-organizations and M-organizations. Let wSi ðx1 ; x2 Þ and wM i ðx1 ; x2 Þ be the worker i’s wage in the S- and M-organizations, respectively. Then, from Eq. (2), we get 1 wSi ðx1 ; x2 Þ ¼ mPðxi Þ 2

14

HIDEO OWAN

wM i ðx1 ; x2 Þ

x 2 xj i ¼ mP 1 P 2 2 3

(7)

The key difference between these two wage functions is that there is no interaction between x1 and x2 in wSi , whereas x1 and x2 are Edgeworth 10 2 S M M This feature drives most of substitutes in wM i (i.e., @ wi =ð@x1 @x2 Þo0). our results that follow.

BASIC RESULTS As a ﬁrst step to see what factors affect the relative efﬁciency of the Sorganization and the M-organization, we compare the production frontiers of the two organizations. By substituting x1 ¼ x2 ¼ x into Eqs. 5 and 6, we obtain the following: x2 41 PðxÞ (8) Y S 4Y M if 1 P 2 The condition requires that the marginal increase of P does not decrease too quickly. When it is satisﬁed, there are gains from specialization. Again, k let us assume PðxÞ ¼ 1 ex . Then, inequality (Eq. (8)) holds if and only if M S k W1 and Y ¼ Y for all x when k ¼ 1. In many results in the chapter, we k assume this speciﬁc function of PðxÞ ¼ 1 ex , with which an increase in k raises the relative efﬁciency of the S-organization. The next proposition investigates how an organizational difference affects the workers’ bargaining power. Proposition 2. 1 wSi ðx; xÞ ¼ ðY S ðx; xÞ Y S ð0; 0ÞÞ 4 1 M m x 2 M wi ðx; xÞ ¼ ðY ðx; xÞ Y M ð0; 0ÞÞ P 4 6 2 The proof is omitted because it is straightforward from Eqs. (4)–(7). In the S-organization, the owner-manager and the workers split the quasirent by half and each worker receives one quarter. Hence, the wage in the S-organization is the same as the union wage if collective bargaining leads to a Nash bargaining solution. In the M-organization, a worker receives less than one quarter of the quasi-rent because of his weaker bargaining power. Namely, in the M-organization, when one worker quits, the output does not

Specialization, Multiskilling, and Allocation of Decision Rights

15

drop as sharply as in the S-organization because of the overlapping investment in the same skill set made by the other worker. Put differently, the loss of human capital caused by the separation is partly offset by the more efﬁcient use of the other worker’s broad skills. Because worker turnover is less costly for the ﬁrm as a result, the M-organization should give its workers lower wages, given the same amount of investment. The difference 2 in the bargaining power is captured by the last term m=6 P x=2 . Propositions similar to this have been proven by Horn and Wolinsky (1988) and Stole and Zwiebel (1996b). They ﬁnd that workers bargaining with the employer gain by forming a union when they are substitutable to one another and lose when they are complementary. In our model, workers in the M-organization is introduced, will receive 14 ðY M ðx; xÞ Y M ð0; 0ÞÞ when collective bargaining 2 and therefore, the union wage is higher by m=6P x=2 in the M-organization. Let x~ S and x~ M be the efﬁcient total investment in the S- and Morganizations, respectively.11 Also, denote the equilibrium investment for the S and M-organizations by xS and xM , respectively. As the owner-manager can capture a share of the surplus created by the workers’ investment (i.e., the holdup problem), the workers underinvest in skills in the S-organization. Namely x~ S 4xS . In the M-organization, however, the investments are substitutes and thus subject to a countervailing effect: a worker’s investment increases his marginal contribution to a subgroup containing only him and the owner more than his marginal contribution to the total output because his broad skills are more valuable without a coworker whose skills overlap with his. A mechanism working behind the Shapley value formula is that a worker’s human capital investment weakens the other worker’s bargaining position, which in turn generates additional negotiation surplus for the former. More intuitively, an increase in one worker’s investment makes it more likely that the other’s discovery of the solution is redundant, reducing the value of the other’s skills. I will call this the ‘‘skill substitution effect.’’ In an extreme case, the workers could overinvest in skills in the M-organization. To see this point, compare the ﬁrst-order conditions for x~ M and xM . The efﬁcient investment x~ M is obtained by solving the following: x x @Y M 1P c0 ðxÞ ¼ 0 ðx; xÞ c0 ðxÞ ¼ mP0 (9) 2 2 @xi xM solves x m x @wM m x i ðx; xÞ c0 ðxÞ ¼ P0 1P þ P0 c0 ðxÞ ¼ 0 3 2 2 6 2 @xi

(10)

16

HIDEO OWAN

Let me rewrite Eq. (10) as follows: x @wM m x m x x i 1P c0 ðxÞ þ P0 P ðx; xÞ c0 ðxÞ ¼ P0 @xi 2 2 2 6 2 2 M M @Y 1 @Y m x x ¼ c0 ðxÞ þ P0 P @xi 2 @xi 6 2 2 |ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ} |ﬄﬄﬄ{zﬄﬄﬄ} Holdup effect

(11)

Skill substitution effect

¼0 In the M-organization, when the skill substitution effect is greater than the holdup effect, the workers overinvest in skills.12 As we discussed in the section on Introduction, Edlin and Hermalin (2000) ﬁnd a similar effect that offsets the holdup effect in a model with substitutable investments made by the buyer and the seller of an asset. Because of the skill substitution effect discussed earlier, the M-organization tends to give the workers stronger incentives to acquire skills if the individual bargaining procedure is adopted. This comparison is more clearly stated when we assume P(x) ¼ 1ex. With this production technology, YS ¼ YM for any symmetric investment x1 ¼ x2 ¼ x. Therefore, x~ S ¼ x~ M . Despite the fact that the wage is lower in the M-organization than in the S-organization, the former outperforms the latter in terms of productivity as shown in the next proposition. Proposition 3. When P(x) ¼ 1ex, xM 4xS . Proof. Suppose x1 ¼ x2 ¼ x. Then, @wM @wS m x x i ðx; xÞ i ðx; xÞ ¼ e2 ð1 e2 Þ40 @xi @xi 6 for all xW0. Hence, xM 4xS . Propositions 2–3 imply that when the two organizational forms have the same production frontier, ﬁrms should choose an M-organization because it induces higher investment and gives the owner-manager a larger share of the economic rent earned.

ALLOCATION OF DECISION RIGHTS In the previous sections, the manager has played only a limited role in determining the ﬁrm’s output because v is constant. In this section, we

Specialization, Multiskilling, and Allocation of Decision Rights

17

assume that the owner-manager allocates decision problems to the workers. In so doing, a manager has to decide how many decision rights she should delegate or, in other words, how much the workers should be ‘‘empowered.’’ Let S be the set of all decision problems. There exists a partition S ¼ S1 , S2 where skill set Ak is necessary to solve problems in Sk. Decision problems in S are ordered according to the relative advantage of the manager to the workers in making right decisions. The decision problems that are high in this ordering are more strategic decisions, and the manager’s ability to coordinate decisions between S1 and S2 gives her relatively high productivity. The decision problems that are low in this ordering are more operational decisions, and the workers’ engagement in the production process enables them to create relatively high value on these lower-order problems. The manager’s problem is simply to choose the set of decision rights she will delegate to the workers. This set is Sr ¼ S1,r , S2,r where r A [0, 1] is the delegation parameter that represents how much the workers are empowered and Sk,r C Sk. We assume that there is a continuum of decision problems and r is continuous. r ¼ 0 indicates that no decision rights are delegated to the workers, and r ¼ 1 means all decision problems are assigned to the workers. Sr is nondecreasing in r in the set order. Let z(r) denote the average incremental value created by a worker from a successful decision on s A Sr and Z(r) denote the average incremental value created by the manager for s A S\Sr. The size of the allocated responsibility affects the quality of implementation and thus the value created by successful decisions: z(r) is decreasing in r, and Z(r) is increasing in r. As before, P(xi,k) is the probability that worker i with investment (xi,1, xi,2) will ﬁnd a right solution for s A Sk,r, and we assume the same properties for P as in the previous sections. The ﬁrm’s expected joint output is as follows: 2 X #ðS k;r Þ½1 ð1 Pðx1;k ÞÞð1 Pðx2;k ÞÞzðrÞ YðX; rÞ ¼ #ðS\Sr ÞZðrÞ þ k¼1

where #(Z) denotes the number of decision problems in Z. Assume #(S1,r) ¼ #(S2,r) and redeﬁne v(r) ¼ #(S\Sr)Z(r) and m(r) ¼ #(Sk,r)z(r). Then, the output function looks similar to Eq. (1) except that the productivity coefﬁcients are parameterized by r as follows: 2 X ½1 ð1 Pðx1;k ÞÞð1 Pðx2;k ÞÞ (12) YðX; rÞ ¼ vðrÞ þ mðrÞ k¼1

18

HIDEO OWAN

We assume that r is not contractible. Otherwise, ex ante transfers can be made contingent on the decision rights delegated to the workers and the ﬁrm can achieve the efﬁcient delegation by simply ‘‘selling’’ decision rights to the workers. Here, the owner-manager is expected to choose r so as to maximize her ex post bargaining surplus. We also assume that once the ownermanager delegates the set of decision rights Sr to the workers, she is unable to solve any problems in S\Sr after the skill investment is made.13 The interpretation is that the owner-manager loses access to the opportunities to learn problem-speciﬁc knowledge and information that the workers acquire after delegation. As the owner-manager cannot take over any subset of delegated decision rights ex post when one or two workers leave the ﬁrm, the proﬁt function is identical to Eq. (3) except for the parameterized coefﬁcients v(r) and m(r). To facilitate mathematical derivation, let v and m be twice continuously differentiable. To ensure unique interior solution, we assume vv r 0, vu(0)W0, vu(1)o0, mv r 0, mu(0)W0, and mu(1)o0. The concavity of v and m also makes it suboptimal for the owner-manager and the workers to share authority and exchange information. Now we ask how increased allocation of decision authority affects the workers’ incentives to acquire skills. Consider the S-organization as before. The workers solve 1 max wSi ðr; x1 ; x2 Þ cðxi Þ ¼ mðrÞPðxi Þ cðxi Þ x 2 From the ﬁrst-order condition 1 mðrÞP0 ðxi Þ ¼ c0 ðxi Þ 2

(13)

dxS m0 ðrÞP0 ðxS Þ ¼ 00 dr 2c ðxS Þ mðrÞP00 ðxS Þ

(14)

we get

Similarly, dxM ¼ dr

x x m0 ðrÞP0 2M 3 2P 2M 2 x x x x 6c00 2M mðrÞP00 2M 32 P 2M þ mðrÞP0 2M

Thus, dxS =dr40 and dxM =dr40 if and only if mu(r)W0.

(15)

Specialization, Multiskilling, and Allocation of Decision Rights

19

The owner-manager will optimize with respect to r, assuming that workers will choose investment optimally. Let pS ðr; xS Þ and let pM ðr; xM Þ denote the proﬁt functions for the owner-manager in the S- and M-organizations, respectively. Then, pS ðr; xS Þ ¼ vðrÞ þ mðrÞPðxS Þ

pM ðr; xM Þ

2 ! xM 1 x ¼ vðrÞ þ 2mðrÞ P P M 2 2 3

vuW0 implies that the owner-manager’s productivity increases as his responsibilities decrease. This means that the owner-manager is overloaded in the range where vuW0. Similarly, the workers are overloaded when mu(r)o0. In Assumption 4, we assume that the owner-manager and the workers cannot be overloaded at the same time. In order words, the set of decision rights S should not be too large for three people to handle. Assumption 4:

If m0 ðrÞ 0; n0 ðrÞo0

The owner-manager solves maxr pl(r) where l ¼ S, M. The ﬁrst-order condition takes the following form: dpl @pl @pl dx ¼ þ l ¼0 dr @r @x dr |{z} |ﬄﬄﬄﬄl{zﬄﬄﬄﬄ} ð1Þ

(16)

ð2Þ

There are two channels through which an increase in r affects the ownermanager’s proﬁt: (1) through a direct impact on productivity and bargaining power and (2) through an indirect impact via a change in worker incentives. Let rl be the optimal degree of delegation for the l-organization. The next lemma is used in the following results. Lemma 1. m0 ðrl Þ40 for l ¼ S,M. Proof. We prove only for the S-organization. The proof is basically the same for the M-organization. dpS 1 1 dx ¼ v0 ðrÞ þ m0 ðrÞPðxS Þ þ mðrÞP0 ðxS Þ S ¼ 0 dr dr 2 2 By substituting in Eqs. (13) and (14), " # dpS 1 0 2 c0 ðxS Þ2 0 ¼ v ðrÞ þ m ðrÞ PðxS Þ þ dr 2 mðrÞ c00 ðxS Þ 12 mðrÞP00 ðxS Þ

(17)

(18)

20

HIDEO OWAN

Suppose m0 ðrS Þ 0. The term in the last brackets is positive. Then, Assumption 4 implies that dpS =drðrM Þo0, leading to contradiction. This concludes the proof. Now we compare the owner-manager’s optimal choice with the efﬁcient allocation r~l that maximizes the total surplus created in the ﬁrm, Y l ðr; xl ðrÞÞ 2cðxl ðrÞÞ. Note that the efﬁcient allocation discussed here does not assume efﬁcient investment in skills. Thus, its efﬁciency is constrained by the workers’ self-interested choice of skill investment. Proposition 4. The owner-manager always underdelegates decision rights to the workers in the S-organization. Underdelegation also appears in the M-organization if the cost function is well-behaved. r~S 4rS When cðXÞ ¼ 12 cx2 ;

r~M 4rM

Proof. See the appendix. Prendergast (1995) similarly argues that managers delegate less responsibilities than would be efﬁcient.14 In his work, underdelegation arises because the ﬁrm or the manager cannot capture all of the intrinsic beneﬁts of implementing tasks (e.g., only the worker acquires skills by ‘‘learning-bydoing’’) in the face of the liquidity constraint imposed on workers. In the framework presented here, the manager underdelegates from the fear of increasing the worker’s bargaining power. From pl ðr; xl ðrÞÞ ¼ Y l ðr; xl ðrÞÞ wl1 ðr; xl ðrÞ; xl ðrÞÞ wl2 ðr; xl ðrÞ; xl ðrÞÞ and @wli =@xi ðxl ðrÞÞ c0 ðxl ðrÞÞ ¼ 0 dpl d @wl @wl @wl dx ðrÞ @wl2 dxl ðrÞ ¼ ðY l ðr; xl ðrÞÞ 2cðxl ðrÞÞÞ 1 2 1 l dr @rﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄ @rﬄ} @x2 dr @x1 dr dr |ﬄﬄﬄﬄﬄﬄﬄﬄ |ﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ{zﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄﬄ ﬄ} Holdup effect

Skill substitution effect

(19) The owner-manager tends to underdelegate because the workers also capture some of the rent directly created by delegation (i.e., @wli =@r40). Let us call this the holdup effect. In the M-organization however there is an offsetting effect: an increase in a worker’s skills in response to larger responsibilities weakens his coworker’s bargaining return (i.e., @wli =@xj dxl ðrÞ=dro0). We call this the skill substitution effect, although it is now used to capture the impact of skill substitution on the ﬁrm owner’s incentive

Specialization, Multiskilling, and Allocation of Decision Rights

21

to delegate decisions, whereas the term was earlier used to refer to the impact on the worker’s incentive to invest in skills. Eq. (19) implies that the distortion in the allocation of decision rights will be smaller in the M-organization unless the skill substitution effect is too large.15 We will now demonstrate that the optimal set of decision rights delegated to the workers tends to be larger in the M-organization. Proposition 5. When P(x) ¼ 1ex, rS orM . Proof. See the appendix. The role of ex post bargaining in getting this result is essential for two reasons. First, it is only under ex post bargaining that the workers with substitutable skills have greater incentive to invest in human capital. As additional delegation tends to induce a greater increase in human capital investment in the M-organization than in the S-organization, the efﬁcient allocation of decision rights is typically more decentralized in the former than in the latter. Second, it is only under ex post bargaining that workers with substitutable skills have weaker bargaining power. Note that the owner-manager usually distorts the allocation of decision rights because she cannot capture all the rents created by the efﬁcient allocation. This is similar to the classic holdup problem. But this underdelegation problem is much less severe in the M-organization where the owner-manager receives a larger share of the efﬁciency gain due to her stronger bargaining power in the M-organization. Under the assumption of identical production frontier, the M-organization gives a higher proﬁt to the owner-manager for two reasons: (1) the owner-manager has stronger bargaining power in the M-organization (Proposition 2) and (2) the workers are better motivated to acquire skills in the M-organization (Proposition 3). More formally, Corollary 1. When P(x) ¼ 1ex, the M-organization gives the owner a higher proﬁt ex post than the S-organization. Proof. pM ðrM ; xM ðrM ÞÞ4pM ðrS ; xM ðrS ÞÞ4pS ðrS ; xM ðrS ÞÞ4pS ðrS ; xS ðrS ÞÞ where the ﬁrst inequality is derived by the fact that rM , is the optimal choice in the M-organization, the second is immediate from Proposition 2 and the third from Proposition 3. It is not clear whether workers’ wages will be higher or lower in the M-organization even with the same assumptions as in Corollary 1. Although

22

HIDEO OWAN

the workers are given more responsibilities and are more motivated to invest in ﬁrm-speciﬁc human capital, which strengthens their bargaining power, they may not receive a higher wage, simply because their share of the rent is smaller in the M-organization (Proposition 2).

COLLECTIVE BARGAINING AND COOPERATION Suppose the workers collectively negotiate their wages by forming a union and thus receive an equal union wage. Horn and Wolinsky (1988) and Stole and Zwiebel (1996b) ﬁnd that workers bargaining with an indispensable ﬁrm gain by forming a union when they are substitutable to one another and lose when they are complementary.16 We restate their result without the proof. The total union wage is deﬁned as the Nash bargaining solution between the owner-manager and the workers, as is assumed in the earlier works. Proposition 6. (Horn & Wolinsky, 1988; Stole & Zwiebel, 1996b) Given the ﬁxed investment by the workers, collective bargaining increases the workers’ wages in the M-organization, while the workers in the S-organization do not beneﬁt from collective bargaining. With collective bargaining, the workers’ share of the rent created by their skills is always half, regardless of the organizational form. In contrast, the workers’ share of the rent obtained through individual bargaining is higher in the S-organization, where the share is half, than in the M-organization, where the share is less than half.17 As you can easily see, collective bargaining reduces the workers’ skill investment because it encourages free-riding. Lower-skill investment also decreases the beneﬁt of delegation for the owner-manager. Hence, the following proposition is straightforward, and thus, the proof is omitted. Proposition 7. Collective bargaining gives a lower incentive to acquire skills and induces the allocation of fewer decision rights to the workers in both organizations. As the workers in the S-organization will be worse off by forming a union, I will focus on the M-organization for now. So far, we have assumed that the workers choose their investment in skills noncooperatively even when they are unionized. When the workers work in teams, they may be able to cooperate (or collude) in choosing their investment level if each worker’s investment is observable to the other. When the workers cooperate, they

Specialization, Multiskilling, and Allocation of Decision Rights

23

jointly optimize their total payoff: w1(x1, x2) þ w2(x1, x2) c(x1) c(x2). We consider the four possible cases: (UC) form a union and cooperate in investment; (NC) not form a union but cooperate in investment; (UN) form a union but not cooperate in investment; and (NN) not form a union and NC UN NN not cooperate in investment. Let xUC M , xM , xM ; and xM be the workers’ skill investments in each of the abovementioned cases. Proposition 8. The workers’ cooperation in setting their investment in skills beneﬁts the owner-manager under collective bargaining but hurts the ﬁrm’s proﬁt under individual bargaining. UC xNN M 4xM 4 max

UN xNC M ; xM

Proof. See the appendix. Cooperation is beneﬁcial under collective bargaining because it eliminates free-riding. Under individual bargaining, however, cooperation and autonomy in setting skill standards in the M-organization could substantially reduce efﬁciency. This is because the strategic substitutability of workers’ investments creates negative externality on their utilities, and internalizing this externality while neglecting the positive externality on the ownermanager’s payoff could cause signiﬁcant underinvestment in skills. This cooperation in skill investment is less likely if there are many teams competing with one another. Also, frequent job rotation among teams working on similar tasks should alleviate this problem because the possibility of moving to another team, or of a worker from another team coming to your team, reduces the incentive to collude. This may partly explain why many Japanese ﬁrms have adopted job rotation.

DISCUSSION When the two organizational forms generate the same production frontier, the ﬁrm should choose the M-organization because it induces higher investment and gives the owner-manager stronger bargaining power. The M-organization is also more efﬁcient unless it induces sufﬁciently excessive investment, which is very unlikely in a team with more than two workers and with sufﬁciently costly investment. This beneﬁt of multiskilling practices has never been recognized in the literature. If there exist gains from

24

HIDEO OWAN k

specialization as we normally expect (e.g., pðxÞ ¼ 1 ex for kW1 in our model), the manager has to make trade-offs between the incentive efﬁciency from multiskilling and the technical efﬁciency from specialization. Eventually, the efﬁcient job design should prevail.18 The difference between the two types of job design becomes remarkable with individual bargaining but not with collective bargaining. Therefore, the M-organization will be favored in workplaces that are not unionized and in those where there is more differentiation in wages based on skill levels. One legitimate critique to our theory may be that in many industries and occupations, especially manufacturing plants, wage bargaining is highly centralized, or a wage formula is typically standardized, even if the plant is not unionized, so that individual bargaining or the Shapley value as the wage bargaining outcome is not realistic. Our reaction to this critique is twofold. First, wages in our model should be interpreted as the lifetime wage income rather than as short-term wages. Even if workers cannot directly bargain over wages in the short run, those with better skills may inﬂuence the managers’ decisions, which affect their income in the long run (e.g., promotion, job assignment, and training). Therefore, the bargaining process formalized in our model captures a worker’s ability to make his supervisors’ decisions reﬂect his voice over a long period of time. Second, although our study was ﬁrst motivated by a recent trend for reorganization toward job enlargement, the application of the framework should not be restricted to the work organization of lower-level workers. For example, the model may be more suitable to explain differences in job design across hierarchical levels. Pay for managers are determined by individual bargaining, much broader cross-functional knowledge/skills are required of managers, and much more decision authority is assigned to managers. Our theory may predict that job design consisting of skill requirement and decision authority assignment should be much closer to the efﬁcient one for managers than for lower-level positions. The model may be also applicable to a comparison among different organizational forms where the relationship between human capital in different units varies. For example, compare M-form and U-form organizations. It may be the case that ﬁrm-speciﬁc human capital across functional units in the U-form organizations are complementary, whereas those across business units in the M-organizations are substitutes if there is a lot of duplication of practices among the units. Then, the model predicts that human capital investment is larger and decision-making is more decentralized in the M-form than in the U-form organizations.

Specialization, Multiskilling, and Allocation of Decision Rights

25

CONCLUSION This work makes an important contribution in two areas. First, it offers another rationale for multiskilling practices increasingly observed in industries worldwide. Multiskilling can reduce the distortion in investment and the allocation of responsibilities created by the holdup problem because multiskilling makes ‘‘bargaining returns’’ more sensitive to skill acquisition and also mitigates the manager’s fear of giving away too much bargaining power to the workers. Therefore, choosing multiskilling practices will enhance productivity unless specialization offers a substantial technical advantage. This idea that multiskilling practices mitigate the distortion in human capital investment and delegation decisions has never been discussed in the literature. Second, we have demonstrated that the multiskilling form of organization could arise as the optimal form even if there are no technological or informational task complementarities among the combined skills. Prior works typically argued that task complementarities are primary reason for many ﬁrms adopting multiskilling practices (e.g., see Lindbeck & Snower, 2000).

NOTES 1. Other terms such as cross-training, multitasking, job enrichment, and job enlargement are similarly used in the literature. Although these words have slightly different meanings, they are used almost interchangeably to describe a recent trend in job design toward a broader set of tasks and responsibilities assigned and a broader set of skills required for an individual job. 2. The way complementarity among tasks determines the degree of specialization has been discussed by authors in the job assignment literature. See Roy (1950), MacDonald (1982), and Rosen (1982) for example. 3. The emphasis on broad skills and decentralization of responsibilities in large Japanese ﬁrms have been also noted by other authors, including Cole (1979), Aoki (1988), Lincoln and Kalleberg (1990), and Kagono, Nonaka, Sakakibara, and Okumura, (1985). 4. See, for example, Arthur (1994), Dunlop and Weil (1996), MacDufﬁe (1995), Ichniowski, Shaw, and Prennushi (1997), Cappelli and Neumark (2001), Hamilton, Nickerson, and Owan (2003), Boning, Ichniowski, and Shaw (2007), Black and Lynch (2001, 2004), Bartel (2004), Kato and Morishima (2002), DeVaro (2006), Eriksson (2003), Bayo-Moriones et al. (2003), and Zwick (2004). 5. One analogy is Bertrand versus Cournot competition. In the former, the sale (wage) is more sensitive to price cuts (human capital investments), but the proﬁt margin (wage level) is lower. 6. The work by Rajan and Zingales (2001) is also relevant to our research. Their work considers the effects of potential market ‘‘competition’’ between individuals

26

HIDEO OWAN

with substitutable human capital on organizational choice, whereas our analysis looks at the effects of within-ﬁrm ‘‘competition’’ among individuals with substitutable human capital. 7. For a more general n worker case, see Owan (1999). 8. Assumption 1 does not ensure that either the S-organization or the Morganization is an efﬁcient form of skill investment. Therefore, it is quite possible that an efﬁcient form is one in which both workers invest in both types of skills, but differentiate their investments in the two types of skills. We need more restrictive conditions to derive the result that either the S-organization or the M-organization is efﬁcient. 9. More formally, the Shapley value is deﬁned as follows:

wi ðXÞ ¼

1 X jSj!ðjNj jSj 1Þ!½YðS [ fig; XÞ YðS; XÞ jNj! SN\fig

where N is the set of all agents including the owner and the workers, S is an arbitrary subset, and Y(S, X) is the expected output by the subgroup S given the investment proﬁle X. 10. It may be often more realistic to assume Edgeworth complementarity between x1 and x2 in the S-organization. Most qualitative results in this chapter hold for such a case. See Owan (1999) for discussions. 11. As we made clear earlier, the current assumptions do not ensure that either the S-organization or the M-organization is efﬁcient. Thus, we mean constrained efﬁciency here: no deviation from the speciﬁed investment proﬁle gives a better payoff to anyone without negatively affecting others within the set of investment proﬁles called the S-organization or the M-organization. 12. For example, suppose cðxi;1 ; xi;2 Þ ¼ 1=2cðxi;1 þ xi;2 Þ2 . Then, the workers overinvest when the cost parameter c is sufﬁciently small. The necessary and sufﬁcient condition for this to be true is pðxM =2Þ43=4. 13. This assumption also implies the owner-manager’s ability to commit to her delegation decision. As ‘‘worker empowerment’’ encourages workers’ skill investment, the owner-manager may promise to delegate large amounts of authority to the workers to motivate them and then take back some of the decision rights to extract more surplus. This assumption rules out such possible reneging by the ownermanager. 14. Freeman and Lazear derive a similar result in a different context where the ﬁrm decides whether it should create a works council to empower workers or not. They argue that even if setting up a works council is efﬁcient, the employer may not be better off by doing so because the empowered workers will likely capture a signiﬁcant part of the efﬁciency gain. 15. There may be some pairs of production functions and cost functions that induce overdelegation in the M-organization, although we have not found any such simple functions. Overdelegation may arise if a greater allocation of decision rights increases the substitutability of the workers’ skills so signiﬁcantly that further delegation weakens the workers’ relative bargaining power. 16. Segal (2003) shows a very general result to this effect concerning the desirability of coalitions under the random-order value, which gives each player his

Specialization, Multiskilling, and Allocation of Decision Rights

27

expected marginal contribution to the set of preceding players in various orderings of players, according to some probability distribution over orderings. The randomorder value in which all orderings are equally likely is the Shapley value. 17. This result may be consistent with the fact that enterprise unions are rare in the United States, where traditional ﬁrms have maintained a specialized job structure. A typical union’s goal of increasing the workers’ bargaining power probably necessitated the use of monopoly power in the industry-wide labor market. On the contrary, in Japan, where large ﬁrms have adopted many multiskilling practices, most unions are enterprise unions. 18. Suppose the owner-manager can commit to a job design at the time of hiring. Then, by adopting the efﬁcient job design, the owner-manager should be able to make the largest proﬁt if she only needs to guarantee young unskilled workers their reservation utility by offering an appropriate upfront wage (i.e., the reservation wage minus the ex post bargaining surplus), or the owner-manager should be able to offer the highest wage if ﬁrms compete for workers in the labor market.

ACKNOWLEDGMENTS I am grateful to Edward P. Lazear, D. John Roberts, Masahiko Aoki, Jeffrey Zwiebel, and Michael Waldman for their encouragement and valuable suggestions. I also thank Jed DeVaro, Hideshi Itoh, Glenn MacDonald, W. Bentley Macleod, Hodaka Morita, Jeroen Swinkels, and Takashi Ui for their useful comments.

REFERENCES Aghion, P., & Tirole, J. (1997). Formal and real authority in organizations. Journal of Political Economy, 105, 1–29. Aoki, M. (1988). Information, incentives, and bargaining in the Japanese economy. Cambridge: Cambridge University Press. Arthur, J. B. (1994). Effects of human resource systems on manufacturing performance and turnover. Academy of Management Journal, 37, 670–687. Bartel, A. P. (2004). Human resource management and organizational performance: Evidence from retail banking. Industrial and Labor Relations Review, 57(2), 181–203. Bayo-Moriones, J. A., Galilea-Salvatierra, P. J., & Merino-Diaz de Cerio, J. (2003). Participation, cooperatives and performance: An analysis of Spanish manufacturing ﬁrms. In: T. Kato & J. Pliskin (Eds.), Determinants of the incidence and the effects of participatory organizations: Advances in the economic analysis of participatory and labormanaged ﬁrms (pp. 31–56). Amsterdam: Elsevier/JAI. Becker, G. S., & Murphy, K. M. (1992). The division of labor, coordination costs, and knowledge. The Quarterly Journal of Economics, 107(4), 1137–1160. Binmore, K. G., Rubinstein, A., & Wolinsky, A. (1986). The Nash bargaining solution in economic modeling. RAND Journal of Economics, 17, 176–188.

28

HIDEO OWAN

Black, S. E., & Lynch, L. M. (2001). How to compete: The impact of workplace practices and information technology on productivity. The Review of Economics and Statistics, 83(3), 434–445. Black, S. E., & Lynch, L. M. (2004). What’s driving the new economy? The beneﬁts of workplace innovation. The Economic Journal, 114(493), F97–F116. Bolton, P., & Dewatripont, M. (1994). The ﬁrm as a communication network. The Quarterly Journal of Economics, 109(4), 809–839. Boning, B., Ichniowski, C., & Shaw, K. (2007). Opportunity counts: Teams and the effectiveness of production incentives. Journal of Labor Economics, 25(4), 613–650. Cappelli, P., & Neumark, D. (2001). Do ‘‘high-performance’’ work practices improve establishment-level outcomes? Industrial and Labor Relations Review, 54(4), 737–775. Cole, R. E. (1979). Work, mobility, & participation. Berkeley, CA: University of California Press. DeVaro, J. (2006). Teams, autonomy, and the ﬁnancial performance of ﬁrms. Industrial Relations: A Journal of Economy & Society, 45(2), 217–269. Dunlop, J. T., & Weil, D. (1996). Diffusion and performance of modular production in the U.S. apparel industry. Industrial Relations: A Journal of Economy & Society, 35, 334–355. Edlin, A. S., & Hermalin, B. E. (2000). Contract renegotiation and options in agency problems. The Journal of Law, Economics, & Organization, 16, 395–423. Eriksson, T. (2003). The effects of new work practices: Evidence from employer-employee data. In: T. Kato & J. Pliskin (Eds.), Determinants of the incidence and the effects of participatory organizations: Advances in the economic analysis of participatory and labormanaged ﬁrms (pp. 3–30). Amsterdam: Elsevier/JAI. Freeman, R., & Lazear, E. (1995). An economic analysis of works councils. In: J. Rogers & W. Streeck (Eds), Works councils: Consultation, representation, and cooperation in industrial relations [NBER conference volume]. Chicago, IL: University of Chicago Press. Grossman, S. J., & Hart, O. D. (1986). The costs and beneﬁts of ownership: A theory of vertical and lateral integration. Journal of Political Economy, 94, 691–719. Hamilton, B. H., Nickerson, J. A., & Owan, H. (2003). Team incentives and worker heterogeneity: An empirical analysis of the impact of teams on productivity and participation. Journal of Political Economy, 111(3), 465–498. Hart, O., & Moore, J. (1990). Property rights and the nature of the ﬁrm. Journal of Political Economy, 98, 1119–1158. Holmstrom, B., & Milgrom, P. (1991). Multitask principal-agent analyses: Incentive contracts, asset ownership and job design. The Journal of Law, Economics, & Organization, 7(Sp.), 24–52. Horn, H., & Wolinsky, A. (1988). Worker substitutability and patterns of unionization. The Economic Journal, 98, 484–497. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The effects of human resource management practices on productivity: A study of steel ﬁnishing lines. The American Economic Review, 87, 291–313. Itoh, H. (1991). Incentives to help in multi-agent situations. Econometrica, 59, 611–636. Itoh, H. (1992). Cooperation in hierarchical organizations: An incentive perspective. The Journal of Law, Economics, & Organization, 8, 321–345. Itoh, H. (1994). Job design, delegation and cooperation: A principal-agent analysis. European Economic Review, 38, 691–700.

Specialization, Multiskilling, and Allocation of Decision Rights

29

Kagono, T., Nonaka, I., Sakakibara, K., & Okumura, A. (1985). Strategic vs. evolutionary management: A U.S.-Japan comparison of strategy and organization. Amsterdam: NorthHolland. Kato, T., & Morishima, M. (2002). The productivity effects of participatory employment practices: Evidence from New Japanese panel data. Industrial Relations: A Journal of Economy & Society, 41(4), 487–520. Koike, K. (1977). Shokuba no Rodo-Kumiai to Sanka (Labor unions and participation in workplaces). Tokyo: Toyo Keizai Shimposha. Koike, K. (1988). Understanding industrial relations in modern Japan. London: Macmillan Press Ltd.. Lincoln, J. R., & Kalleberg, A. L. (1990). Culture, control, and commitment. Cambridge, NY: Cambridge University Press. Lindbeck, A., & Snower, D. J. (2000). Multitask learning and the reorganization of work: From tayloristic to holistic organization. Journal of Labor Economics, 18(3), 353–376. MacDonald, G. M. (1982). A market equilibrium theory of job assignment and sequential accumulation of information. The American Economic Review, 72, 1038–1055. Macdufﬁe, J. P. (1995). Human resource bundles and manufacturing performance: Organizational logic and ﬂexible production systems in the world auto industry. Industrial and Labor Relations Review, 48, 197–221. Milgrom, P., & Roberts, J. (1990). The economics of modern manufacturing: Technology, strategy, and organization. The American Economic Review, 80, 511–528. Milgrom, P., & Roberts, J. (1995). Complementarities and ﬁt: Strategy, structure, and organizational change in manufacturing. Journal of Accounting and Economics, 19, 179–208. Owan, H. (1999). Internal organization, bargaining and human capital. Dissertation, Stanford University. Rosen, S. (1982). Authority, control, and the distribution of earnings. Bell Journal of Economics, 13, 311–323. Roy, A. D. (1950). The distribution of earnings and of individual output. The Economic Journal, 60, 489–505. Prendergast, C. J. (1995). A theory of responsibility in organizations. Journal of Labor Economics, 13, 387–400. Rajan, R. G., & Zingales, L. (2001). The ﬁrm as a dedicated hierarchy: A theory of the origins and growth of ﬁrms. The Quarterly Journal of Economics, 116, 805–851. Segal, I. (2003). Collusion, exclusion, and inclusion in random-order bargaining. The Review of Economic Studies, 70(2), 439–460. Stole, L. A., & Zwiebel, J. (1996a). Intra-ﬁrm bargaining under non-binding contracts. The Review of Economic Studies, 63, 375–410. Stole, L. A., & Zwiebel, J. (1996b). Organizational design and technology choice under intraﬁrm bargaining. The American Economic Review, 86, 195–222. Zwick, T. (2004). Employee participation and productivity. Labour Economics, 11(6), 715–740.

30

HIDEO OWAN

APPENDIX Proof of Proposition 1 We consider four cases: (1) xi;k 40 for all i and k; (2) xi;1 ¼ 0 but xi;2 40 for all i; (3) xi;1 ¼ xi;2 ¼ 0 for all i; and (4) xi;1 ¼ xj;2 ¼ 0 but xi;2 40 and xj;1 40 without loss of generality. Case 1 In this case, all ﬁrst-order conditions hold with equality and m 0 m P ðxi;1 Þð3 2Pðxj;1 ÞÞ ¼ P0 ðxi;2 Þð3 2Pðxj;2 ÞÞ ¼ c0 ðxi;1 þ xi;2 Þ 6 6 m 0 m (A.1) P ðxj;1 Þð3 2Pðxi;1 ÞÞ ¼ P0 ðxj;2 Þð3 2Pðxi;2 ÞÞ ¼ c0 ðxj;1 þ xj;2 Þ 6 6 If m=6P0 ðxi;1 Þð3 2Pðxj;1 ÞÞ4m=6P0 ðxj;1 Þð3 2Pðxi;1 ÞÞ, then m=6P0 ðxi;2 Þð3 2Pðxj;2 ÞÞ4m=6P0 ðxj;2 Þð3 2Pðxi;2 ÞÞ from Eqs. (A.1) and Assumption 1 implies xi;1 oxj;1 and xi;2 oxj;2 . But they contradict c0 ðxi;1 þ xi;2 Þ4c0 ðxj;1 þ xj;2 Þ, the other implication from Eq. (A.1). Similarly, assuming m=6P0 ðxi;1 Þð3 2Pðxj;1 ÞÞom=6P0 ðxj;1 Þð3 2Pðxi;1 ÞÞ leads to contradiction. Therefore, m=6P0 ðxi;1 Þð3 2Pðxj;1 ÞÞ ¼ m=6m=6P0 ðxj;1 Þð3 2Pðxi;1 ÞÞ. Then m=6P0 ðxi;2 Þð3 2Pðxj;2 ÞÞ ¼ m=6P0 ðxj;2 Þð3 2Pðxi;2 ÞÞ from (A.1) and Assumption 1 implies xi;1 ¼ xj;1 and xi;2 ¼ xj;2 . If xi;1 ¼ xj;1 oxi;2 ¼ xj;2 , m=6P0 ðxi;1 Þð3 2Pðxj;1 ÞÞ4 m=6 P0 ðxi;2 Þð3 2Pðxj;2 ÞÞ contradicting Eq. (A.1). Hence, xi;1 ¼ xj;1 ¼ xi;2 ¼ xj;2 . It is easily seen that there exists the unique interior solution for Eq. A.1 from Assumption 3 and the concavity of the objective function. Case 2 The ﬁrst-order conditions are m 0 m P ð0Þð3 2Pðxj;1 ÞÞ P0 ðxi;2 Þð3 2Pðxj;2 ÞÞ ¼ c0 ðxi;2 Þ 6 6 m 0 m 0 P ðxj;1 Þð3 2Pð0ÞÞ P ðxj;2 Þð3 2Pðxi;2 ÞÞ ¼ c0 ðxj;1 þ xj;2 Þ 6 6

(A.2)

If m=6 P0 ðxi;2 Þð3 2Pðxj;2 ÞÞ4m=6 P0 ðxj;2 Þð3 2Pðxi;2 ÞÞ, Assumption 1 suggests xi;2 oxj;2 . But Eq. (A.2) also indicates c0 ðxi;2 Þ4c0 ðxj;1 þ xj;2 Þ contradicting xi;2 oxj;2 . Thus, m=6P0 ðxi;2 Þð3 2Pðxj;2 ÞÞ m=6 P0 ðxj;2 Þð3 2Pðxi;2 ÞÞ. If xj;1 40, this combined with Eq. (A.2) implies m=6 P0 ð0Þð3 2Pðxj;1 ÞÞ

Specialization, Multiskilling, and Allocation of Decision Rights

31

m=6 P0 ðxj;1 Þð3 2Pð0ÞÞ, which suggests xj;1 ¼ 0 from Assumption 1. As m=6 P0 ð0Þð3 2Pð0ÞÞ4m=6 P0 ðxÞð3 2PðxÞÞ for any x, this result contradicts Eq. (A.2). Case 3 The ﬁrst-order conditions are m 0 m 0 P ð0Þð3 2Pðxj;1 ÞÞ c0 ð0Þ; P ð0Þð3 2Pðxj;2 ÞÞ c0 ð0Þ 6 6 m 0 P ðxj;k Þð3 2Pð0ÞÞ ¼ c0 ðxj;1 þ xj;2 Þ if xj;k 40 6

(A.3)

If xj;k 40; c0 ðxj;1 þ xj;2 Þ c0 ð0Þ and m=6 P0 ðxj;k Þð3 2Pð0ÞÞ4m=6 P0 ð0Þð3 2Pðxj;1 ÞÞ: The last inequality contradicts Assumption 1. Therefore, xi;1 ¼ xj;1 ¼ xi;2 ¼ xj;2 ¼ 0. But this case is ruled out by Assumption 2. Case 4 The ﬁrst-order conditions are m 0 m P ð0Þð3 2Pðxj;1 ÞÞ P0 ðxi;2 Þð3 2Pð0ÞÞ ¼ c0 ðxi;2 Þ 6 6 m 0 m 0 P ð0Þð3 2Pðxi;2 ÞÞ P ðxj;1 Þð3 2Pð0ÞÞ ¼ c0 ðxj;1 Þ 6 6

(A.4)

Eq. (A.4) indicates xi;2 ¼ xj;1 . Therefore, there exists an equilibrium when m=6 P0 ð0Þð3 2Pðx ÞÞ c0 ðx Þ for the solution x for m=6 P0 ðx Þ ð3 2Pð0ÞÞ ¼ c0 ðx Þ.

Proof of Proposition 4 Let r be the number such that m0 ðrÞ ¼ 0. Then, mu(r)W0 if and only if 0oror. From Lemma 1, 0orl or. r~l solves l dY l dpl @wl1 @wl2 @w1 @wl1 @wl2 @wl2 dxl ¼ þ þ þ ¼0 þ þ þ dr dr @r @r @x1 @x2 @x1 @x2 dr For 8ror, @wli =@r40 and dxl =dr40 from Eqs. (14) and (15) and mu(r)W0. For the S-organization, @wSi =@xi 40 and @wSi =@xj ¼ 0 holds additionally. Therefore, dY S =dr4dpS =dr for 8ror. This implies that dY S =dr4dpS =dr 0 for 8r rS leading to r~S 4rS . For the M-organization,

32

HIDEO OWAN

@wM @wM 1 x 2 x x i þ i ¼ mðrÞP0 M mðrÞP0 M P M @xi @xj 2 2 2 2 3 x 1 2 x 40 P M ¼ mðrÞP0 M 2 2 2 3 if and only if PðxM =2Þo3=4. Hence, it is not clear if dY M =dr4dpM =dr holds for 8ror in general. Now assume cðXÞ ¼ 1=2 cx2 : Then, x x m0 ðrÞP0 2M 3 2P 2M 2 x x x x 6c00 2M mðrÞP00 2M 32 P 2M þ mðrÞP0 2M x m0 ðrÞ c0 2M m0 ðrÞ xM o xM ¼ mðrÞ c00 mðrÞ 2

dxM ¼ dr

(A.5)

2

where the inequality is derived from the ﬁrst-order condition mðrÞP0 ðxM =2Þð1=2 1=3PðxM =2ÞÞ ¼ c0 ðxM Þ and Pvo0. M M To show that @wM i =@r þ ð@wi =@xi þ @wi =@xj ÞdxM =dr40 always holds, we assume PðxM =2Þ43=4. Then, M @wM @wi @wM dxM xM 2 x 0 i i þ 4m ðrÞP 1 P M þ @r @xi @xj dr 2 2 3 x x 1 2 x 0 M 0 M P P M þ m ðrÞ 2 2 2 2 3 from Eq. (A.5) xM 2 xM xM 1 2 x 0 0 4m ðrÞP 1 P þ m ðrÞP P M 2 2 2 2 3 2 3 x 3 4 x ¼ m0 ðrÞP M 40 P M 2 2 2 3 where the second inequality is obtained from P(x)WxPu(x), a property of concave functions with P(0) ¼ 0 and Pu(x)W0. This leads to dY M =dr4dpM =dr for 0 8ror implying that dY M =dr40 for 0 8r rM . Therefore, r~M 4rM .

33

Specialization, Multiskilling, and Allocation of Decision Rights

Proof of Proposition 5 By substituting P(x) ¼ 1ex into Eq. (18), we get " # dpS 1 1 c0 ðxS Þ2 0 0 ¼ v ðrÞ þ m ðrÞ PðxS Þ þ dr 2 mðrÞ c00 ðxS Þ 12 mðrÞP00 ðxS Þ " # 1 1 c0 ðxS Þ2 0 0 ¼ v ðrÞ þ m ðrÞ PðxS Þ þ 2 mðrÞ c00 ðxS Þ þ 12 mðrÞexS 1 c0 ðxS Þ 1 ¼ v0 ðrÞ þ m0 ðrÞ PðxS Þ þ 2 mðrÞ c00 ðxS Þ=c0 ðxS Þ þ 1

ðA:6Þ

where the last line is derived by using the ﬁrst-order condition @wSi 1 1 ¼ mðrÞP0 ðxS Þ ¼ mðrÞexS ¼ c0 ðxS Þ @xi 2 2 Similarly,

" 2 dpM x 2 x 0 0 ¼ v ðrÞ þ m ðrÞ 2P M P M dr 2 2 3 3 þ

2 mðrÞ

2 ¼ v0 ðrÞ þ m0 ðrÞ "

4 2P

x 4v ðrÞ þ m ðrÞ 2P M 2 0

0

ðxM Þ2

7 c 2 7 5 x x x c00 ðxM Þ 12 mðrÞ P00 2M 12 13 P 2M 13 P0 2M 0

xM 2

2 x 2 23 P 2M þ mðrÞ

c0 ðxM Þ2 c00 ðxM ÞþmðrÞ

x M 1 xM 1 2 þ12 e 3e

3 5

# 2 2 xM c0 ðxM Þ 1 P (A.7) þ 2 3 mðrÞ c00 ðxM Þ=c0 ðxM Þ þ 1

where the last line is derived by using the ﬁrst-order condition @wM mðrÞ 0 xM xM mðrÞ 0 xM i 1P þ P P ¼ @xi 2 2 2 3 6 mðrÞ x mðrÞ xM ¼ e Mþ e 2 ¼ c0 ðxM Þ 3 6

34

HIDEO OWAN

Compare Eqs. (A.6) and (A.7) term by term. From xM 4xS , the result in Proposition 3, 2 1 1 1 x 4 1 xM 2 x xM 2 x 2 M M o þ e e ¼ 2P P M PðxS Þo PðxM Þ ¼ e 2 2 2 2 2 3 3 3 3 c0 ðxS Þoc0 ðxM Þ; and 1 c00 ðxS Þ=c0 ðxS Þ

þ1

o

1 c00 ðxM Þ=c0 ðxM Þ

þ1

where the last inequality is derived from the assumption that cv(x)/cu(x) is nonincreasing. dpM S Hence, dp dr o dr for all r such that mu(r)W0. This concludes the proof.

THE EFFECT OF MULTISKILLING ON LABOR PRODUCTIVITY, PRODUCT QUALITY, AND FINANCIAL PERFORMANCE Martin Farnham and Emma Hutchinson ABSTRACT A substantial literature examines the effect of high-performance workplace practices on various outcomes for ﬁrms and workers. However, little attention has been paid to the effect of broad job design on product quality or ﬁnancial performance. And with rare exception, the empirical literature on outcomes from high-performance work practices treats those practices as exogenously determined. This chapter seeks to address these two shortcomings in the existing literature. Using a nationally representative cross-section of British employers in 2004, we measure the effect of multiskilling on establishment-level labor productivity, product quality, and ﬁnancial performance. We ﬁnd that treating multiskilling as an endogenous choice of employers in empirical models of organizational performance has signiﬁcant implications for the results. In particular, the estimated (positive) effect of multiskilling on labor productivity vanishes when we treat multiskilling as an endogenous choice of employers. Treating multiskilling as an endogenous choice changes its estimated

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 35–62 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012006

35

36

MARTIN FARNHAM AND EMMA HUTCHINSON

effect on product quality from zero to positive and substantially increases the estimated magnitude of its (positive) effect on ﬁnancial performance. Keywords: Training; job design JEL classiﬁcations: J24; M53

INTRODUCTION Over the past three decades, researchers in business, sociology, and economics have investigated the causes and consequences of various ‘‘high-performance work practices’’ among manufacturers. Many such practices arose in Japanese manufacturing and were adopted by Western manufacturers beginning in the 1980s. Multiskilling – training workers to perform a variety of tasks – is one of these practices. In this chapter, we investigate the effect of multiskilling on organizational performance, as measured by labor productivity, product quality, and ﬁnancial performance. In an important departure from the existing empirical literature on high-performance workplace practices and their effects on organizational outcomes, we treat multiskilling as an endogenously chosen variable in our analysis. We show that treating the choice to multiskill workers as endogenous rather than exogenous substantially changes the empirical estimates we obtain. Our ﬁndings, on their own, provide a useful contribution to the literature on the effects of multiskilling. Very little empirical work has been done to date on the effects of multiskilling (or even broad job design more generally) on organizational performance. However, our ﬁndings also have broader implications for the study of other high-performance work practices (e.g., teamwork, quality circles, total quality management, multitasking, and job rotation). The vast majority of the empirical literature on the effects of highperformance work practices treats workplace practices as exogenously determined. Our ﬁndings, therefore, suggest the need for a more careful empirical approach in this large, broader literature. Multiskilling is an element of ‘‘broad job design,’’ a common feature of modern human resource practice. Broad job design includes multitasking – where employees perform multiple tasks throughout their work shift rather than specializing in one task and job rotation – where employees perform one task or set of tasks for a period of time, then switch to another task or

Effect of Multiskilling

37

set of tasks for a period of time. Because multiskilling (knowing how to do multiple tasks) is both a prerequisite to and a consequence of multitasking and job rotation, multiskilling is an integral feature of broad job design. The effect of multiskilling on organizational performance is of interest for several reasons. Over the past two decades, ﬁrms have spent billions of dollars reengineering their organizational practices to improve efﬁciency and, ultimately, proﬁtability. Therefore, understanding the effect of practices such as multiskilling is clearly of interest to business. It is also of interest to designers of job training programs, who must decide whether to specialize or multiskill their trainees in preparation for reentry into the work force. Furthermore, the effects of multiskilling are of interest to those who design educational curricula. Some national educational systems value specialization, often at an early age, whereas others value broader training. If multiskilling is shown to have beneﬁcial effects for business, this may induce educational systems to place greater emphasis on preparing their students for careers in a multiskilled workforce. We study the effect of multiskilling on ﬁrm performance using the 2004 British Workplace Employment Relations Survey (WERS). The 2004 WERS is a nationally representative cross-section of establishments including both manufacturers and nonmanufacturers. We measure the effect of multiskilling on ﬁrm performance using two basic approaches. In the ﬁrst approach, we estimate three different single-equation probit models with multiskilling on the right-hand side, using as dependent variables measures of labor productivity, product quality, and ﬁnancial performance of the establishment relative to its industry. Because multiskilling is a choice made by employers, this approach – which assumes multiskilling is exogenously determined – is likely to produce estimates subject to selectivity bias. Providing evidence in support of this concern, Nickell, Nicolitsas, and Patterson (2001) use British data to show that poorly performing ﬁrms are more likely to reform workplace practices than are high-performing ﬁrms. This suggests that estimates of the effect of multiskilling on our organizational performance outcomes are likely to be negatively biased. Although the possibility of selectivity bias is largely ignored in the existing empirical literature on workplace practices, another potential source of bias comes in the form of time-invariant unobserved heterogeneity that is correlated with workplace practices and affects organizational performance. For example, suppose that multiskilling is more easily adopted with a more educated workforce. Such a workforce may also be more productive. Failure to control for human capital levels of the workforce could therefore lead to upward bias in estimates of the effect of multiskilling on labor productivity.

38

MARTIN FARNHAM AND EMMA HUTCHINSON

A third potential source of bias is the fact that organizations adopting innovative work practices may be engaged in other, unobserved, activities that enhance performance. In this case, one would expect estimates of the effects of multiskilling to be positively biased. Finally, a fourth potential source of bias is error in measures of organizational performance that is correlated with workplace practice. These different sources of endogeneity could collectively introduce bias of unknown sign into estimates generated under the assumption of exogeneity of workplace practices. Therefore, in our second approach to estimating the effect of multiskilling on organizational performance, we explicitly address the endogeneity of multiskilling by estimating a bivariate probit for each measure of performance and by treating multiskilling as an endogenous right-hand side variable in the organizational performance equations. In each of the bivariate probit models, the multiskilling equation includes an exogenous variable that reﬂects volatility in the product market and is excluded from the performance equation. DeVaro and Farnham (2011) demonstrate theoretically and empirically why product market volatility is a determinant of an employer’s decision to multiskill its workers. It should be noted that, while this exclusion restriction is a source of identiﬁcation in our bivariate probit models, identiﬁcation does not hinge on this assumption. This is because, in our speciﬁcations, the bivariate probit is identiﬁed without an exclusion restriction. In the absence of an exclusion restriction, identiﬁcation comes from the nonlinearity of the bivariate probit. When we treat multiskilling as exogenous, we ﬁnd that multiskilling is associated with higher labor productivity and higher ﬁnancial performance but is unrelated to product quality. When we treat multiskilling as endogenous, our results change. In this case, we ﬁnd that multiskilling is unrelated to labor productivity – a ﬁnding at odds with much of the current literature on broad job design – but is associated with higher ﬁnancial performance and higher product quality. Many high-performance workplace practices that were introduced in manufacturing have been adopted by ﬁrms outside the manufacturing sector. Yet, the empirical literature on the effect of workplace practices has been slow to shift its focus from manufacturing. Therefore, this study, through its use of a nationally representative sample of all establishments, serves a further useful purpose in helping to extend the empirical job design literature to focus on workplace practices among nonmanufacturing ﬁrms, which make up the vast majority of economic activity in Western economies.

39

Effect of Multiskilling

LITERATURE REVIEW A number of studies seek to explain the rise of broad job design over recent decades. Lindbeck and Snower (2000) describe the transition of manufacturers from ‘‘Tayloristic’’ to ‘‘holistic’’ organizations, with Tayloristic ﬁrms characterized by high levels of specialization and repetition by workers and holistic ﬁrms characterized by job rotation, integration of tasks, and learning across tasks. Lindbeck and Snower argue that the shift to multitask learning, led by Japanese manufacturers, has resulted from the rise of computers and information systems, increased ﬂexibility of the capital stock, the growth of human capital per worker, and changes in worker tastes in favor of a more varied work experience. Boucekkine and Crifo (2008) extend the model to a dynamic framework and endogenize worker skills. Several studies address the impact of multiskilling on worker incentives. Owan (2011) argues that multiskilling creates task overlap and makes worker skills substitutable, thus creating competition among workers and providing them an incentive to acquire ﬁrm-speciﬁc human capital. Carmichael and MacLeod (1993) write that multiskilled workers may have greater incentives than specialized workers to reveal hidden information about the optimal implementation of technological change in the ﬁrm’s production process. Cosgel and Miceli (1999) argue that multiskilling enhances peer monitoring (Kandel & Lazear, 1992) by giving workers information about each other’s jobs. Other studies highlight the role of learning by workers and employers that results from broad job design. Ortega (2001) and Eriksson and Ortega (2006) argue that employers who face imperfect information use job rotation to learn which workers are best at which tasks. This allows for more proﬁtable matching of workers to positions. A different set of studies note that multiskilled workers are essential to process improvement. Koike (1985) and Aoki (1986) note that multiskilled workers are better than specialists at dealing with emergent events (such as equipment breakdown) on the production line. Wang (2002) argues that multiskilling aids coordination of workers. Morita (2005) and Gibbs, Levenson, and Zoghi (2010) argue that workers with a broad understanding of production can better suggest improvements. Although much of the literature on broad job design has focused on its role in facilitating process innovation, DeVaro and Farnham (2011) argue that specialized workers are better than multiskilled workers at product innovation. They model a multiproduct ﬁrm that can train workers to produce multiple products or specialize workers in a single product. In the

40

MARTIN FARNHAM AND EMMA HUTCHINSON

model, beneﬁts of multiskilling come in the form of increased ﬂexibility in labor allocation that allows resources to be quickly shifted from less proﬁtable to more proﬁtable goods in the ﬁrm’s product line. Costs of multiskilling come in the form of a diminished ability to alter a good to meet shifting consumer demand due to broad but shallow training. More volatility in the product market means ﬁrms need to innovate more to produce goods that meet the current speciﬁcations demanded by consumers. DeVaro and Farnham’s model predicts that greater product market volatility will lead to greater specialization of workers. Using data from the 2004 British WERS, they corroborate their theoretical prediction with empirical results that increased product market volatility leads to less multiskilling. Their ﬁnding forms the basis for the exclusion restriction used in this chapter and discussed further in the section on Empirical Strategy and Analysis. One of the empirical questions we address in this chapter is the relationship between multiskilling and labor productivity. Of the abovementioned theoretical studies, several imply a positive effect of broad job design on labor productivity through improved incentives, better matching, or better understanding of the production process. On the contrary, economists since Adam Smith (1776) have pointed to productivity gains from specialization. Therefore, theoretical claims about the effect of multiskilling on labor productivity are mixed and arguably the sign of this effect remains an empirical question. Ichniowski, Shaw, and Prennushi (1997) use longitudinal data on steel ﬁnishing lines to measure the effect of human resource management systems and individual workplace practices on plant performance. In ﬁxed effects models including only job rotation as the work practice of interest, they ﬁnd job rotation to be positively associated with productivity. However, when they include controls for human resource management systems, the coefﬁcient on job rotation becomes negative. Cappelli and Neumark (2001) use longitudinal US data from the National Employers Survey and the Longitudinal Research Database to investigate the effect of high-performance work practices on labor productivity and labor costs. Under the assumption that no high-performance work practices were in place in 1977, they ﬁnd that while high-performance practices as a group generally raise labor productivity, job rotation tends to be associated with lower labor productivity. Although these studies control for unobserved time-invariant heterogeneity by use of ﬁxed effects methods, neither of them addresses the potential selectivity bias resulting from the fact that job design is an endogenous choice of managers and neither directly addresses the effect of multiskilling on labor productivity.

Effect of Multiskilling

41

The only study of broad job design that treats workplace practice as an endogenous choice of managers is Zwick (2002). He treats continuous job training as endogenous and includes job rotation on the right-hand side in his model. He does not directly address the endogeneity of job rotation, however. Zwick ﬁnds no effect of job rotation on ﬁrm productivity, using a panel of German ﬁrms. A number of more general studies of high-performance work practices address the effect of bundles of practices on labor productivity. Most ﬁnd high-performance work practices to be positively related to labor productivity (Caroli & Van Reenen, 2001; Huselid, 1995; Ichniowski et al., 1997). Black and Lynch (2001, 2004) ﬁnd that only certain combinations of such practices are associated with higher productivity. Huselid and Becker (1996) ﬁnd no effect of such practices on labor productivity once controlling for unobserved ﬁrm-level heterogeneity. Another empirical question we address is the effect of multiskilling on product quality. Osterman (1994) argues that ﬁrms pursuing ‘‘high road’’ practices such as offering high-quality products are more likely to engage in high-performance work practices. Wang (2002) directly addresses multiskilling when he assumes that multiskilling will increase product quality by increasing coordination between work groups and reducing production errors. Morita (2005) also suggests that product quality will beneﬁt from multiskilling of workers due to its contribution to the improved functioning of quality circles. To our knowledge, no empirical study has addressed the impact of multiskilling (or even broad job design) on product quality. Although Ichniowski et al. (1997) ﬁnd that bundles of high-performance work practices are associated with higher product quality, they do not speciﬁcally address the empirical relationship between product quality and broad job design. Many of the theoretical papers mentioned earlier also have implications for the effect of multiskilling on ﬁnancial performance. Ceteris paribus, increases to labor productivity or product quality should lead to greater revenues and hence greater proﬁts for a ﬁrm. However, multiskilling may also result in higher training costs. And multiskilling could affect ﬁrm ﬁnancial performance through other channels such as the removal of layers of management. Although economists generally assume ﬁrms engage in activities that contribute to proﬁt maximization, it remains possible that some persistent management trends are counterproductive. Although Huselid (1995) and Huselid and Becker (1996) discuss effects on ﬁnancial performance, they draw inference from data on labor costs and labor productivity, rather than data on proﬁts or other measures of overall

42

MARTIN FARNHAM AND EMMA HUTCHINSON

ﬁnancial performance. To our knowledge, the speciﬁc relationship between multiskilling and ﬁnancial performance has not been previously studied. Importantly, only one of the aforementioned empirical studies (Zwick, 2002) treats work practices as an endogenous choice of organizations and therefore addresses selectivity bias. DeVaro (2008) provides a list of 24 articles on the effects of workplace practices on ﬁrm and worker outcomes, all but one of which treat workplace practices as exogenously determined.1 Although several studies use panel data to control for time-invariant unobserved heterogeneity at the ﬁrm or establishment level (e.g., Black & Lynch, 2001, 2004; Cappelli & Neumark, 2001; Caroli & Van Reenen, 2001; Huselid & Becker, 1996; Ichniowski et al., 1997), such ﬁxed effects approaches cannot control for time-varying heterogeneity, which is likely to underlie the differential choice of organizations whether or not to engage in practices like multiskilling. Without controlling for selection into treatment, these studies risk providing biased estimates of the effect of the treatment on organizational performance. Our study contributes to the literature on job design and highperformance work practices in the following ways. Ours is the ﬁrst paper, to our knowledge, to investigate the effect of multiskilling on product quality or ﬁnancial performance. Perhaps more importantly, we treat multiskilling as an endogenous choice of organizations and show that addressing potential selectivity bias makes a difference in the estimates we obtain. This allows us to make a new contribution to the small literature on broad job design and labor productivity. Furthermore, it illustrates that in the broader literature on the effects of workplace practices – including practices and outcomes not included in our study2 – there is reason to treat the endogenous choice of workplace practices carefully. Our paper also contributes usefully to the literature by employing a nationally representative sample of establishments from both manufacturing and nonmanufacturing sectors. Therefore, the results we present come not from a single ﬁrm, industry, or sector, but from the British economy as a whole and are hence more likely to be generalizable than results of more narrow studies.

DATA Our data source is the management questionnaire from the 2004 British WERS, jointly sponsored by the Department of Trade and Industry, the Advisory, Conciliation and Arbitration Service, the Economic and Social Research Council, and the Policy Studies Institute. Distributed through the

43

Effect of Multiskilling

UK Data Archive in November 2005, the WERS data are a nationally representative stratiﬁed random sample covering British workplaces with at least ﬁve to nine employees, except for local units in Northern Ireland and those in the following 2003 Standard Industrial Classiﬁcation (SIC) divisions: agriculture, hunting, and forestry; ﬁshing; mining and quarrying; private households with employed persons; and extraterritorial organizations. The 2004 WERS was the ﬁfth such survey, following earlier waves in 1980, 1984, 1990, and 1998. The sampling frame used for WERS 2004 is the Inter-Departmental Business Register (IDBR), which is maintained by the Ofﬁce for National Statistics (ONS). As noted by Chaplin, Mangla, Purdon, and Airey (2005), ‘‘The IDBR is undoubtedly the highest quality sample frame of organisations and establishments in Britain. The frame is continuously up-dated from [administrative tax] records and establishments that no longer exist are removed reasonably quickly.’’ Our measure of multiskilling is based on the answer to the following question: ‘‘Approximately, what proportion of [workers in the establishment’s largest occupational group] are formally trained to be able to do jobs other than their own?’’ From this, we deﬁne the following binary response variable: Multiskilling ¼ 1 if the proportion is positive ¼ 0 if the proportion is zero The respondent for each establishment in the WERS is asked how (1) the current labor productivity of the establishment; (2) the current product quality of the establishment; and (3) the current ﬁnancial performance of the establishment compare with other establishments in the same industry. In answering each question, respondents can choose between ‘‘A lot better than average,’’ ‘‘Better than average,’’ ‘‘About average for industry,’’ ‘‘Below average,’’ ‘‘A lot below average,’’ and ‘‘No comparison possible.’’ On the basis of these responses, we construct three binary measures of organizational performance as dependent variables in our statistical models: Labor Productivity, Product Quality, and Financial Performance. Each of these variables equals 1 if the establishment’s current performance relative to that of other establishments in the industry is reported to be above average and 0 if it is reported to be at or below average. The survey also includes a question about how the respondent interprets the term ‘‘ﬁnancial performance,’’ with possible responses including proﬁt, costs, sales, value added, and so on. We show that restricting the sample for the ﬁnancial performance models to those employers interpreting the

44

MARTIN FARNHAM AND EMMA HUTCHINSON

variable to mean ‘‘proﬁt’’ makes little difference in the estimates obtained. Therefore, we use the full sample for most of our analysis. One potential criticism of our methodology is the fact that we use a dataset with subjective performance measures. In defense of these measures, the WERS is an anonymous survey and so respondents do not have strong incentives to give misleading answers. Reporting errors may be a problem, although the respondent is typically a senior manager at the establishment and therefore is presumably well informed about the performance of the establishment relative to the industry. And, as evidenced by recent accounting scandals, reporting problems can also occur with objective measures of organizational performance such as accounting proﬁts. There is evidence suggestive of systematic reporting errors in the subjective performance measures, as seen by the fact that more than half of respondents report their establishment is doing better than the industry average (Table 1). The skewness of the distribution of performance outcomes becomes more apparent when you consider that the percentage of respondents who say that their establishment is performing below the industry average is only 7%, 2%, and 8% for Labor Productivity, Product Quality, and Financial Performance, respectively. DeVaro (2008) discusses this issue. One explanation for such a skewed distribution of reported performance is that establishments existing at the time of the survey are more successful than the average establishment in the population. Unsuccessful ﬁrms tend to go out of business quickly and are therefore less likely to be captured in a survey than successful ﬁrms. Another possibility is that managers may tend to overreport performance out of a sense of pride in their work. One can interpret this as a relabeling of the categories describing organizational performance. As long as this relabeling is uncorrelated with the treatment, it poses no problem for our estimates. Of greatest concern is the possibility that some systematic reporting error is correlated with the decision to multiskill workers. For instance, more optimistic managers may tend to be more receptive to high-performance work practices while also tending to be overly bullish on the performance of the establishment.3 Because we use an estimation strategy that explicitly accounts for correlation between unobserved determinants of multiskilling and organizational performance, systematic reporting errors that are correlated with the treatment should not confound our estimates. Summary measures of descriptive statistics for all variables in the analysis are displayed in Table 1. We use establishment sampling weights when computing the statistics in this table and in all of our subsequent analyses. Table 1 displays descriptive statistics for the variables in analysis and

45

Effect of Multiskilling

Table 1.

Descriptive Statistics for 2004 WERS Sample. Mean

Standard Error

Multiskilling ( ¼ 1 if some workers multiskilled)

0.602

0.018

Labor Productivity ( ¼ 1 if greater than industry avg.) Product Quality ( ¼ 1 if greater than industry avg.) Financial Performance ( ¼ 1 if greater than ind. avg.)

0.515 0.791 0.500

0.019 0.015 0.019

Volatility ( ¼ 1 if product market ‘‘turbulent’’)

0.169

0.015

2.690 0.295 0.329 0.021 14.475 0.392 0.269 0.101 0.892 0.025 5.303 0.218 0.119

0.025 0.015 0.011 0.004 0.918 0.024 0.017 0.012 0.012 0.005 0.584 0.013 0.009

Industry: Manufacturing Electricity, gas, and water Construction Wholesale and retail Hotels and restaurants Transport and communication Financial services Other business services Public administration Education Health Other community services

0.111 0.001 0.049 0.250 0.089 0.049 0.052 0.148 0.022 0.049 0.116 0.065

0.013 0.0002 0.008 0.016 0.010 0.008 0.008 0.013 0.004 0.005 0.010 0.008

Largest occupational group at workplace: Professionals Associate professional and technical Administrative and secretarial occupations Skilled trades Caring, leisure, and personal service Sales and customer service

0.072 0.087 0.144 0.098 0.097 0.236

0.008 0.009 0.012 0.012 0.010 0.016

Log of establishment size Union Fraction of part-time workers Fraction of temporary workers Percent union Number of recognized unions Owner-manager Foreign owned Establishment is at least 5 years old Franchise Fixed-term percentage Fixed term Temporary workers

46

MARTIN FARNHAM AND EMMA HUTCHINSON

Table 1. (Continued )

Process, plant, and machine operatives and drivers Routine unskilled Sample size ¼ 2,295

Mean

Standard Error

0.117 0.150

0.012 0.013

Notes: Statistics are computed on the full sample. Due to missing values, the sample size varies across variables; all nonmissing observations were used to compute each statistic. Summary statistics based on the analysis subsample are available upon request and are similar to those reported here.

describes the distribution of establishments in the sample across industry and occupation of core production workers. Some of the variables contain missing values, and we estimate all of our models using listwise deletion.4

EMPIRICAL STRATEGY AND ANALYSIS We use and compare two basic empirical strategies in the analysis that follows. The ﬁrst strategy involves estimating simple single-equation probits with a measure of organizational performance – Labor Productivity, Product Quality, or Financial Performance – as the dependent variable and Multiskilling as the key independent variable. In this approach, we assume that the dependent variable of interest is determined by an unobserved latent variable Y i ¼ X 0 i b þ MS0 i d þ ui

(1)

where i ¼ 1,y, N indexes establishments. Yi is observed and Yi ¼ 1 if the unobserved Y i 0. Yi ¼ 0 otherwise. MSi is a dummy variable indicating whether the establishment employs multiskilling in the workplace. Xi is a vector of establishment-level characteristics including measures of establishment size and age, unionization, management structure, characteristics of worker contracts, and foreign ownership status. These control variables are detailed in the appendix. We do not include industry controls, as the dependent variable is deﬁned relative to the industry mean. ui is an error term, which we assume to be iid normal. Given this assumption, our data are described by a probit model, PrðY i ¼ 1Þ ¼ FðMS0 i d þ X 0 i bÞ where F denotes the cumulative normal distribution function.

(2)

47

Effect of Multiskilling

Underlying this speciﬁcation is the implicit assumption that Multiskilling is exogenously determined. This approach is consistent with the vast majority of the existing literature on workplace practices. Results obtained from this approach are given in Table 2. Given that Multiskilling is an establishment-level choice, and therefore may be correlated with the error term, our second, and preferred, empirical strategy treats this variable as endogenously determined. We do this by estimating bivariate probits with Multiskilling and one of our three organizational performance variables treated as the two endogenous variables. As in Eq. (1), we assume that Multiskilling is determined by an unobserved latent variable, MSi ¼ V 0i p þ X 0 i g þ vi

(3)

where the observed MSi is equal to 1 if MSi 0 and is equal to 0 otherwise. Again, i indexes establishments, and Xi is the same vector of establishment-level characteristics as in Eq. (1). Vi denotes an exogenous determinant of Multiskilling described later. We assume that ui and vi are jointly normally distributed with mean zero and variance one, and we let r ¼ Corr(ui, vi). Recent work by DeVaro and Farnham (2011) shows theoretically and empirically why volatility in the product market is a determinant of an employer’s decision to multiskill its workers. Following that analysis, we deﬁne a binary measure of product market volatility. Volatility ¼ 1 if the current state of the market for the main product or service of the establishment is described as ‘‘turbulent’’ (¼ 0 otherwise).5 To aid in identiﬁcation, we include in Eq. (3) the exogenous dummy variable, V, denoting Volatility. V is excluded from the organizational performance equation (Eq. (1)). Excluding Volatility from the organizational performance equation is appropriate if Volatility only affects performance through the Multiskilling variable. This is arguably a reasonable assumption. Although product market volatility is likely to affect the variance of labor productivity, product quality, and ﬁnancial performance, it does not seem likely to affect the mean of any of these measures of organizational performance. For example, in an industry with high product market volatility, the probability of consumers demanding a product that is very easy to make or very hard to make will be higher than in an industry with low volatility. But the mean of labor productivity should not vary with product market volatility. This is

48

MARTIN FARNHAM AND EMMA HUTCHINSON

Table 2. Multiskilling and Organizational Performance Relative to Own Industry (Coefﬁcient Estimates from Single-Equation Probits).

Multiskilling Mean marginal effect Log establishment size Union Fraction of part-time workers Fraction of temporary workers Percent union Number of recognized unions Owner-manager Foreign owned Establishment is at least 5 years old Franchise Fixed-term percentage Fixed term Temporary workers Constant Sample size

Labor Productivity

Product Quality

Financial Performance

0.484** (0.168) 0.179 0.022 (0.073) 0.111 (0.314) 0.439 (0.273) 0.339 (1.52) 0.001 (0.005) 0.273 (0.215) 0.102 (0.185) 0.124 (0.245) 0.267 (0.267) 0.253 (0.398) 0.001 (0.004) 0.001 (0.004) 0.080 (0.248) 0.265 (0.355) 749

0.158 (0.186) 0.037 0.181 (0.089) 0.132 (0.367) 0.476 (0.303) 0.141 (0.213) 0.004 (0.005) 0.356 (0.230) 0.700 (0.222) 0.394 (0.293) 0.492 (0.275) 1.335 (0.439) 0.003 (0.005) 0.016 (0.254) 0.323 (0.255) 0.458 (0.402) 793

0.520** (0.166) 0.198 0.123 (0.075) 0.089 (0.287) 0.183 (0.272) 1.564 (1.773) 0.001 (0.005) 0.170 (0.188) 0.184 (0.180) 0.065 (0.245) 0.085 (0.262) 0.245 (0.471) 0.0002 (0.0045) 0.135 (0.211) 0.412 (0.254) 0.602* (0.357) 765

Notes: Unless otherwise stated, cell entries give the coefﬁcient estimates from single-equation probits with standard errors in parentheses beneath each estimate. The dependent variables (Labor Productivity, Product Quality, and Financial Performance) are each a dummy variable equal to 1 if the respondent reports that the establishment is performing above the industry average and equal to 0 if it is performing at or below the industry average. **Statistical signiﬁcance at the 5% level for a two-sided alternative. *Statistical signiﬁcance at the 10% level.

Effect of Multiskilling

49

especially true given that our organizational performance measures are deﬁned relative to the industry mean. One might argue that attrition will be higher in industries with greater product market volatility, as a result of higher variance in ﬁnancial performance. If survey respondents do not take this attrition into account when stating how their organization performs relative to the industry, this could lead to ﬁrms in high-volatility industries reporting higher mean values of organizational performance relative to the industry mean than those in low-volatility industries. One might expect, however, that survey respondents would be familiar enough with their industry to take attrition into account when reporting on organizational performance relative to the industry mean. Even if one takes issue with the validity of our exclusion restriction, identiﬁcation does not hinge on this assumption. In contrast to the linear simultaneous equations model, the parameters in the bivariate probit model with a dummy endogenous variable on the right-hand side are identiﬁed (except in unusual cases, such as one demonstrated in Maddala 1983, that do not apply here) even in the absence of exclusion restrictions (Heckman, 1978; Monfardini & Radice, 2008; Wilde, 2000). Identiﬁcation comes from the nonlinearity of the bivariate probit. Given that the model can be estimated with and without the exclusion restriction, we are able to test the restriction. Following DeVaro and Farnham (2011), we restrict our analysis to multiproduct establishments. Both the theory and the empirical evidence in DeVaro and Farnham (2011) suggest restricting the sample to multiproduct establishments.6 In Table 2, we report estimation results from single-equation probit models using as dependent variables Labor Productivity (column 1), Product Quality (column 2), and Financial Performance (column 3). This approach, in which Multiskilling is treated as exogenous, is standard in the literature. Focusing on results that are statistically signiﬁcant at the 5% level using two-tailed tests, the main result from Table 2 is that, ceteris paribus, multiskilling is positively associated with both labor productivity and ﬁnancial performance but not product quality. The magnitudes of the effects in the labor productivity and ﬁnancial performance models are similar and substantial. In both cases, the incremental effect of changing Multiskilling from zero to one on the probability that the dependent variable equals 1 is about 0.2, which amounts to a 35–40% increase in the probability that the establishment is performing above the industry average for those outcomes.7

50

MARTIN FARNHAM AND EMMA HUTCHINSON

Tables 3 and 4 display the main results of our chapter. We report the results of bivariate probit estimation for the labor productivity model (Table 3, columns 1 and 2), the product quality model (Table 3, columns 3 and 4), and the ﬁnancial performance model (Table 4, columns 1 and 2). In each of the three models, Multiskilling appears as an endogenous right-hand side variable in the equation for organizational performance, and the crossequation error correlation, r, is treated as a free parameter. The bivariate probit results given in Tables 3 and 4 differ markedly from the single-equation probit results of Table 2. In the bivariate labor productivity model (Table 3, columns 1–2), the estimated effect of Multiskilling is now statistically insigniﬁcant, in contrast to the singleequation probit model of Table 2 where it was positive and signiﬁcant. And in the product quality model (Table 3, columns 3–4), the estimated effect is now positive and signiﬁcant, in contrast to the single-equation probit model of Table 2 where it was insigniﬁcant. Only for ﬁnancial performance is the qualitative result the same between the two tables, namely that multiskilling is associated with an increased probability of high ﬁnancial performance. And even in this case, the magnitude of the mean incremental effect of changing Multiskilling from 0 to 1 is nearly three times what it was in the single-equation speciﬁcation. The Volatility coefﬁcient in the Multiskilling equation is estimated with low precision; in the ﬁnancial performance model, it is statistically insigniﬁcant, and in the labor productivity and product quality models, it is signiﬁcant only at the 10% level and 5% level, respectively, for one-tailed tests. Another result of interest from Tables 3 and 4 concerns the estimated value of r, the cross-equation disturbance correlation. For both the product quality and the ﬁnancial performance models, r is negative, large in magnitude, and estimated with high precision. For the labor productivity model, r is 0.43 but statistically insigniﬁcant. An important result that can be seen by comparing Table 2 (where r ¼ 0 is implicitly imposed) and Tables 3 and 4 (where r is a free parameter to be estimated) is that the endogeneity of Multiskilling matters a lot for measuring its effect on organizational performance. The Multiskilling coefﬁcient is highly sensitive to the value of r, and the pattern of results reveals that lower values of r are associated with higher values of the coefﬁcient on Multiskilling. Given that r is so low in the ﬁnancial performance and product quality models, the estimated coefﬁcient of Multiskilling is large in those models, and in the case of product quality, it becomes statistically signiﬁcant whereas it was insignificant when r was constrained to be zero in Table 2. Similarly, because r ¼ 0.42 in the labor productivity model, the estimated Multiskilling

Table 3. Multiskilling and Organizational Performance Relative to Own Industry–Labor Productivity, and Product Quality (Coefﬁcient Estimates from Bivariate Probits). Labor Productivity

Multiskilling Mean marginal effect Volatility Log establishment size Union Fraction of part-time workers Fraction of temporary workers Percent union Number of recognized unions Owner-manager Foreign owned Establishment is at least 5 years old Franchise Fixed-term percentage Fixed term Temporary workers Constant r Sample size

Productivity

Multiskilling

0.238 (1.044) 0.091 –

–

0.289 (0.204) 0.332** (0.092) 0.451 (0.349) 0.295 (0.305) 0.095 (1.66) 0.013 (0.008) 0.639 (0.419) 0.260 (0.191) 0.458 (0.289) 0.419** (0.289) 0.839** (0.519) 0.002 (0.005) 0.355 (0.259) 0.399 (0.290) 0.318 (0.378)

0.045 (0.119) 0.164 (0.331) 0.311 (0.321) 0.287 (1.43) 0.001 (0.006) 0.192 (0.217) 0.151 (0.199) 0.010 (0.305) 0.163 (0.304) 0.457 (0.449) 0.002 (0.005) 0.226 (0.225) 0.029 (0.264) 0.037 (0.569) 0.431 (0.578) 745

Product Quality Quality 1.402** (0.304) 0.443 –

Multiskilling –

0.314* (0.175) 0.038 0.334** (0.089) (0.085) 0.220 0.406 (0.295) (0.326) 0.281 0.228 (0.288) (0.283) 0.049 0.256 (0.216) (0.189) 0.0001 0.011** (0.0042) (0.005) 0.415** 0.500 (0.198) (0.332) 0.498** 0.182 (0.208) (0.180) 0.159 0.394 (0.284) (0.257) 0.539 0.316 (0.260) (0.270) 0.816 0.551 (0.401) (0.444) 0.002 0.348 (0.004) (0.241) 0.083 0.348 (0.251) (0.241) 0.079 0.440* (0.231) (0.235) 0.942** 0.367 (0.365) (0.367) 0.768** (0.332) 788

Notes: Unless otherwise stated, cell entries give the coefﬁcient estimates from bivariate probits with standard errors in parentheses beneath each estimate. The dependent variables (Labor Productivity and Product Quality) are each a dummy variable equal to 1 if the respondent reports that the establishment is performing above the industry average and equal to 0 if it is performing at or below the industry average. The endogenous variable (Multiskilling) is a dummy variable equal to 1 if the respondent reports that at least some of the core production employees at the establishment are multiskilled. **Statistical signiﬁcance at the 5% level for a two-sided alternative. *Statistical signiﬁcance at the 10% level.

52

MARTIN FARNHAM AND EMMA HUTCHINSON

coefﬁcient is considerably smaller (i.e., negative and insigniﬁcant) compared with the corresponding coefﬁcient in Table 2. The result that Multiskilling appears to have no effect on labor productivity in the bivariate probit should be interpreted with caution, because the parameter of primary interest (i.e., the Multiskilling coefﬁcient) is highly sensitive to the value of r, which is estimated with very low precision in the case of labor productivity. While r ¼ 0.43, it is statistically insigniﬁcantly different from zero, and as seen in Table 2, if r ¼ 0, then Multiskilling has a clear positive association with labor productivity that is large in magnitude. The sharp differences between Tables 2 and Tables 3–4 reveal that the endogeneity of Multiskilling matters a lot, but the exclusion restriction that contributes to identiﬁcation of the bivariate probit models might be questioned and merits further scrutiny. As noted earlier, the bivariate probit model is, in principle, identiﬁed even in the absence of exclusion restrictions. We reestimate the models of Table 3, including Volatility in all equations. These results (available from the authors upon request) are qualitatively identical to those we report in Tables 3–4 and quantitatively extremely similar. Furthermore, in the organizational performance equations, the Volatility coefﬁcient is far from statistically signiﬁcant (z ¼ 0.13 in the labor productivity model, z ¼ 0.24 in the product quality model, and z ¼ 0.96 in the ﬁnancial performance model). In the Multiskilling equations, where the theoretical model in DeVaro and Farnham (2011) predicts a negative effect of volatility, precision is modest; yet, the Volatility coefﬁcient is in all cases statistically signiﬁcant at the 10% level using one-tailed tests. In summary, the statistical results are supportive of the exclusion of Volatility from the organizational performance models, and in any event, our main results do not hinge on this identifying assumption and hold even in its absence. Our results concerning ﬁnancial performance are of particular interest, given that the literature on broad job design has tended to focus on other outcomes (e.g., labor productivity) that are intermediate to ﬁnancial performance. Financial performance, or proﬁt, is of interest because it is the broadest possible measure of organizational performance, incorporating the full spectrum of beneﬁts and costs of multiskilling. But the term ‘‘ﬁnancial performance’’ is subject to multiple interpretations, and although the modal response in the WERS data interprets the term as synonymous with proﬁt, there are also other interpretations. For this reason, for the ﬁnancial performance model, we reestimate the bivariate probit (from Table 4, columns 1–2) using the subsample of employers interpreting ﬁnancial performance as synonymous with proﬁt. The results, displayed in Table 4, columns 3–4, are qualitatively the same as those in Table 4, columns 1–2,

53

Effect of Multiskilling

Table 4. Multiskilling and Financial Performance Relative to Own Industry (Coefﬁcient Estimates from Bivariate Probits). Full Analysis Sample

Financial Performance Multiskilling Mean marginal effect Volatility Log establishment size Union Fraction of part-time workers Fraction of temporary workers Percent union Number of recognized unions Owner-manager Foreign owned Establishment is at least 5 years old Franchise Fixed-term percentage Fixed term Temporary workers

1.773** (0.114) 0.531 –

Multiskilling

–

Subset of Analysis Sample that Equates Financial Performance with Proﬁts Financial Performance 1.740** (0.149) 0.534 –

Multiskilling

–

0.071 (0.072) 0.134 (0.246) 0.021

0.162 (0.155) 0.353** (0.086) 0.370 (0.336) 0.290

0.079 (0.101) 0.215 (0.299) 0.441

0.089 (0.205) 0.280** (0.114) 0.508 (0.372) 0.506

(0.243) 1.261

(0.289) 0.585

(0.308) 0.389

(0.372) 3.898

(1.751) 0.006 (0.005) 0.299*

(2.36) 0.015** (0.005) 0.316

(1.910) 0.013** (0.006) 0.009

(2.512) 0.017** (0.007) 0.271

(0.179) 0.276* (0.155) 0.205 (0.212) 0.269

(0.312) 0.287 (0.181) 0.453* (0.250) 0.393

(0.215) 0.305 (0.189) 0.075 (0.269) 0.202

(0.445) 0.216 (0.224) 0.375 (0.335) 0.297

(0.219) 0.184 (0.372) 0.001 (0.004) 0.242 (0.194) 0.116 (0.248)

(0.276) 0.604 (0.439) 0.003 (0.004) 0.442* (0.233) 0.521* (0.305)

(0.256) 0.523* (0.298) 0.0004 (0.0045) 0.359 (0.298) 0.251 (0.307)

(0.371) 0.825* (0.448) 0.001 (0.005) 0.245 (0.340) 0.432 (0.374)

54

MARTIN FARNHAM AND EMMA HUTCHINSON

Table 4. (Continued ) Full Analysis Sample

Financial Performance Constant r Sample size

0.982** (0.306)

Multiskilling

0.418 (0.359) 0.967** (0.036) 761

Subset of Analysis Sample that Equates Financial Performance with Proﬁts Financial Performance

Multiskilling

0.720** 0.303 (0.352) (0.449) 0.945** (0.065) 464

Notes: Unless otherwise stated, cell entries give the coefﬁcient estimates from bivariate probits with standard errors in parentheses beneath each estimate. The dependent variable, Financial Performance, is a dummy variable equal to 1 if the respondent reports that the establishment is performing above the industry average and equal to 0 if it is performing at or below the industry average. The endogenous variable (Multiskilling) is a dummy variable equal to 1 if the respondent reports that at least some of the core production employees at the establishment are multiskilled. **Statistical signiﬁcance at the 5% level for a two-sided alternative. *Statistical signiﬁcance at the 10% level.

although the sample size is much smaller. We perform a similar comparison across samples for the single equation probit from Table 2, column 3, and again, results do not change substantively.8

DISCUSSION A substantial literature exists examining the effect of high-performance workplace practices on various outcomes for ﬁrms and workers. However, little attention has been paid to the effect of broad job design on product quality or ﬁnancial performance. And almost without exception, the empirical literature on outcomes from high-performance work practices has treated those practices as exogenously determined.9 This chapter seeks to address these two shortcomings in the existing literature. We estimate the effect of multiskilling on three measures of ﬁrm performance – labor productivity, product quality, and ﬁnancial performance. We begin with an estimation strategy that assumes multiskilling is exogenously determined. We then estimate bivariate probits that treat multiskilling as an endogenous choice of managers. Results change qualitatively for the labor

Effect of Multiskilling

55

productivity and product quality models when we treat multiskilling as endogenous. And while results for the ﬁnancial performance model remain qualitatively the same as in the single-equation probit, the magnitude of the estimated effect of multiskilling is much higher when multiskilling is treated as endogenous. Tests of our identifying restriction suggest that it is valid. Results obtained without the exclusion restriction – identifying off of the nonlinear functional form – are qualitatively identical and quantitatively very similar. In labor productivity models, we ﬁnd that when we treat multiskilling as exogenously determined, it has a positive and statistically signiﬁcant relationship with labor productivity, but when we treat multiskilling as endogenously determined, it has no statistically signiﬁcant relationship with labor productivity. These ﬁndings are consistent with bias resulting from a form of reverse causality. Broad job design is generally thought to require higher levels of human capital than is narrow job design (specializing workers). For example, Lindbeck and Snower (2000) argue that the rise of multitasking is partly attributable to rising human capital levels in the workforce. Therefore, the positive association between multiskilling and labor productivity that we ﬁnd in our single-equation probit may simply result from the fact that ﬁrms with more skilled (and hence productive) workers are more likely to adopt multiskilling. It is worth noting however that our estimate of r, the correlation of the error terms in the Labor Productivity and Multiskilling models, is very imprecise. Therefore, the difference in estimates between the single-equation probit and bivariate probit should not be overemphasized in this case. Our ﬁndings on labor productivity contribute to a small literature of mixed empirical ﬁndings on broad job design and productivity. Ichniowski et al. (1997) ﬁnd that job rotation raises labor productivity when other work practices are not controlled for. This ﬁnding is consistent with our estimate from the single-equation probit. They ﬁnd a negative effect of job rotation on labor productivity when controlling for other work practices. Zwick (2002) ﬁnds no effect of job rotation on productivity. We ﬁnd no effect of multiskilling, when controlling for endogeneity of multiskilling. As noted earlier, a more general literature that looks at the effect of bundles of work practices on labor productivity (Black & Lynch, 2001, 2004; Caroli & Van Reenen, 2001; Huselid, 1995; Ichniowski et al., 1997) tends to ﬁnd positive effects. Given the imprecision with which we estimate r, it is difﬁcult to argue that we can resolve the mixed ﬁndings in the literature. In product quality models, we ﬁnd that when we treat multiskilling as exogenous, it has no statistically signiﬁcant relationship with product quality, but when we treat multiskilling as endogenously chosen, it has a statistically signiﬁcant positive relationship with product quality. This is

56

MARTIN FARNHAM AND EMMA HUTCHINSON

consistent with poorly performing ﬁrms adopting reforms to change course (Nickell et al., 2001). If ﬁrms with product quality problems are more likely to adopt multiskilling, this will tend to bias downward the effect of multiskilling on quality in a single-equation probit speciﬁcation. In this case, the estimate of r we obtain from the bivariate probit is large in magnitude and precisely estimated, suggesting that selectivity bias is a serious problem in the product quality models. In ﬁnancial performance models, we ﬁnd that when we treat multiskilling as exogenously determined, it has a statistically signiﬁcant positive relationship with ﬁnancial performance, but when we treat multiskilling as endogenously chosen, the statistically signiﬁcant positive relationship nearly triples in size. This too is consistent with Nickell et al. (2001) and the selectivity bias that we would expect to follow from those ﬁndings. The estimate of r that we obtain from the bivariate probit is large in magnitude and precisely estimated, suggesting that selectivity bias is a serious problem in the ﬁnancial performance models. This is true whether we use the full analysis sample or restrict the sample to establishments whose respondent manager equates ﬁnancial performance with proﬁt. We use product market volatility as an excluded variable in the bivariate probit speciﬁcations on grounds that product market volatility should affect the choice of multiskilling (DeVaro & Farnham, 2011) but should not affect labor productivity, product quality, or ﬁnancial performance. Volatility is signiﬁcant at the 10% level for a one-sided test in each of the bivariate multiskilling probits. As this exclusion restriction is not required for identiﬁcation, we estimate separate bivariate probits without the restriction and obtain qualitatively identical results. It is worth noting some ways in which our approach in this chapter is limited. The nature of our dataset restricts us to a cross-sectional analysis. Although the WERS has a panel component to it, the measure of product market volatility that we use as an identifying restriction is only available in the 2004 survey.10 Therefore while we are able to estimate the error correlation in the bivariate probits, we cannot ascertain anything about how this relationship has changed over time. Another shortcoming of our approach is that our use of qualitative responses in place of organizational performance measures limits, somewhat, the applicability of our ﬁndings. We cannot convert our ﬁndings into a dollar return to multiskilling, for instance. We discuss potential problems with qualitative responses in greater detail in the section on Data. The main disadvantage of our approach is that we do not analyze systems of work practices. Holmstrom and Milgrom (1994) argue that workplace

Effect of Multiskilling

57

practices need to be analyzed together, not in isolation. Although a number of empirical studies have attempted to study workplace practices in various bundles, the empirical evidence in support of this approach is mixed. Cappelli and Neumark (2001), for instance, ﬁnd little evidence of important clustering effects of workplace practices. Huselid (1995) also ﬁnds limited evidence for important clustering effects. Black and Lynch (2001) ﬁnd that some combinations of workplace characteristics have important interactions, but these mostly involve whether or not the establishment is unionized. Black and Lynch (2004) argue that one is more likely to discover such synergies in intra-industry studies than in representative samples of ﬁrms, due to a greater ability to control for unobserved heterogeneity. Because there exists little theoretical direction on how to specify interactions between workplace practices, we study multiskilling in isolation. If important synergies between multiskilling and other workplace practices do exist, this could potentially bias our estimates of the effect of multiskilling on organizational outcomes. With those caveats, our approach presents some important advantages relative to the existing literature. We explicitly address the selectivity bias resulting from endogenous choice of workplace practices, which has only rarely been done in this literature. We measure relationships between multiskilling and two organizational outcomes – product quality and ﬁnancial performance – that have not been previously measured in the literature. Furthermore, compared to industry studies (Ichniowski et al., 1997) or studies of the manufacturing sector (e.g., Black & Lynch, 2001, 2004), we beneﬁt from the generalizability that comes from using a representative sample of British establishments. Manufacturers make up just 13% of the WERS, and given the increasing application of high-performance work practices in nonmanufacturing organizations, it is useful to broaden the empirical focus of this literature to include these organizations. In future work, we plan to estimate the effect of multiskilling using data with accounting measures of productivity and ﬁnancial performance. Such quantitative measures of the impact of multiskilling would arguably be of greater use than the qualitative measures we obtain here. We also plan to explore the mechanism by which ﬁnancial performance improves when multiskilling is employed. Our ﬁndings that multiskilling has no effect on labor productivity but is associated with higher product quality and better ﬁnancial performance are plausibly consistent with each other. But it would be useful to shed more light on how multiskilling ultimately improves ﬁnancial performance. Measuring effects of multiskilling on other outcomes, such as labor costs, could lend added insight.

58

MARTIN FARNHAM AND EMMA HUTCHINSON

NOTES 1. DeVaro (2006) is the exception. 2. For instance Godard (2001) and Askenazy, Caroli, and Marcus (2002) investigate the effect of workplace practices on worker well-being, rather than organizational performance. 3. DeVaro (2008). 4. In Table 1, we use all available observations for each variable to compute summary statistics. Analogous tables that compute summary statistics based on the smaller subsamples on which the multivariate statistical models are estimated closely match Table 1. 5. Other possible responses were ‘‘growing,’’ ‘‘mature,’’ or ‘‘declining.’’ 6. The model in DeVaro and Farnham is of a ﬁrm deciding whether to train workers in the production of one good or multiple goods. Thus it technically only applies to a multiproduct ﬁrm. Empirical results corroborating their theory are only found in the subsample of multiproduct ﬁrms. 7. Our data contain a categorical measure of multiskilling that denotes different degrees of multiskilling. In results not given here, we include this richer measure of multiskilling as a set of dummies on the right-hand side of our probit models. Results are quantitatively very similar and qualitatively identical to those obtained using a binary measure of multiskilling. For clarity of exposition, we choose to report results in Tables 2–4 using just the binary measure. 8. Results available upon request. 9. Zwick (2002), which treats continuous training as endogenously chosen, and DeVaro (2006) and DeVaro (2008), which treat teamwork as an endogenous choice of managers, are exceptions to this rule. 10. Huselid and Becker (1996) note that use of panel data, especially over short periods, can introduce bias from measurement error that more than outweighs the bias reduction resulting from panel methods that control for establishment-level heterogeneity.

ACKNOWLEDGMENTS The authors acknowledge the Department of Trade and Industry, the Economic and Social Research Council, the Advisory, Conciliation and Arbitration Service and the Policy Studies Institute as the originators of the 2004 Workplace Employment Relations Survey data, and the Data Archive at the University of Essex as the distributor of the data. None of these organizations bears any responsibility for the authors’ analysis and interpretations of the data. The authors would like to thank an anonymous referee and numerous colleagues for helpful feedback.

59

Effect of Multiskilling

REFERENCES Aoki, M. (1986). Horizontal vs. vertical information structure of the ﬁrm. The American Economic Review, 76(5), 971–983. Askenazy, P., Caroli, E., & Marcus, V. (2002). New organizational practices and working conditions: Evidence from France in the 1990s. Louvain Economic Review, 68(1–2), 91–110. Black, S., & Lynch, L. (2001). How to compete: The impact of workplace practices and information technology on productivity. Review of Economics and Statistics, 83(3), 434–445. Black, S., & Lynch, L. (2004). What’s driving the new economy? The beneﬁts of workplace innovation. Economic Journal, 114(493), 97–116. Boucekkine, R., & Crifo, P. (2008). Human capital accumulation and the transition from specialization to multitasking. Macroeconomic Dynamics, 12, 320–344. Cappelli, P., & Neumark, D. (2001). Do ‘high performance’ work practices improve establishment-level outcomes. Industrial and Labor Relations Review, 54(4), 737–775. Carmichael, H., & MacLeod, W. (1993). Multiskilling, technical change and the Japanese ﬁrm. The Economic Journal, 103, 142–160. Caroli, E., & Van Reenen, J. (2001). Skilled biased organizational change? Evidence from a panel of British and French establishments. Quarterly Journal of Economics, 116(4), 1449–1492. Chaplin, J., Mangla, J., Purdon, S., & Airey, C. (2005). The workplace employment relations survey (WERS) 2004 technical report (cross-section and panel surveys). Prepared for Department of Trade and Industry, November. Cosgel, M., & Miceli, T. (1999). Job rotation: cost, beneﬁts, and stylized facts. Journal of Institutional and Theoretical Economics, 155, 301–320. DeVaro, J. (2006). Teams, autonomy, and the ﬁnancial performance of ﬁrms. Industrial Relations: A Journal of Economy & Society, 45(2), 217–269. DeVaro, J. (2008). The effects of self-managed and closely managed teams on labor productivity and product quality: An empirical analysis of a cross-section of establishments. Industrial Relations: A Journal of Economy & Society, 47(4), 659–697. DeVaro, J., & Farnham, M. (2011). Two perspectives on multiskilling and product market volatility. Labour Economics (forthcoming). Eriksson, T., & Ortega, J. (2006). The adoption of job rotation: Testing the theories. Industrial and Labor Relations Review, 59(4), 653–666. Gibbs, M., Levenson, A., & Zoghi, C. (2010). Why are jobs designed the way they are? Research in Labor Economics, 30, 107–154. Godard, J. (2001). High performance and the transformation of work? The implications of alternative work practices for the experience and outcomes of work. Industrial and Labor Relations Review, 54(4), 776–805. Heckman, J. (1978). Dummy endogenous variables in a simultaneous equation system. Econometrica, 46(4), 931–959. Holmstrom, B., & Milgrom, P. (1994). The ﬁrm as an incentive system. American Economic Review, 84(4), 972–991. Huselid, M. (1995). The impact of human resource management practices on turnover, productivity, and corporate ﬁnancial performance. Academy of Management Journal, 38(3), 635–872.

60

MARTIN FARNHAM AND EMMA HUTCHINSON

Huselid, M., & Becker, B. (1996). Methodological issues in cross-sectional and panel estimates of the human resource–ﬁrm performance link. Industrial Relations: A Journal of Economy & Society, 35(3), 400–422. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The effects of human resource management practices on productivity: A study of steel ﬁnishing lines. American Economic Review, 87(3), 291–313. Kandel, E., & Lazear, E. (1992). Peer pressure and partnerships. Journal of Political Economy, 100(4), 801–817. Koike, K. (1985). Human resource development in the Japanese industry. Manuscript. Lindbeck, A., & Snower, D. (2000). Multitask learning and the reorganisation of work: From Tayloristic to Holistic organization. Journal of Labor Economics, 18(3), 353–376. Maddala, G. S. (1983). Limited-dependent and qualitative variables in econometrics. Cambridge, MA: Cambridge University Press. Monfardini, C., & Radice, R. (2008). Testing exogeneity in the bivariate probit model: A Monte Carlo study. Oxford Bulletin of Economics and Statistics, 70(2), 271–282. Morita, H. (2005). Multi-skilling, delegation and continuous process improvement: A comparative analysis of US-Japanese work organizations. Economica, 72, 69–93. Nickell, S., Nicolitsas, D., & Patterson, M. (2001). Does doing badly encourage management innovation? Oxford Bulletin of Economics and Statistics, 63(1), 5–28. Ortega, J. (2001). Job rotation as a learning mechanism. Management Science, 47(10), 1361–1370. Osterman, P. (1994). How common is workplace transformation and who adopts it? Industrial & Labor Relations Review, 47(2), 173–188. Owan, H. (2011). Specialization, multiskilling and allocation of decision rights. Advances in the Economic Analysis of Participatory and Labor-Managed Firms, 12 (forthcoming). Smith, A. (1776). An inquiry into the nature and causes of the wealth of nations. Reprint. Edited by Edwin Cannan. Chicago: University of Chicago Press, 1976. Wang, Y. (2002). Product market conditions and job design. HRRI Working Paper no. 04-02. Industrial Relations Center, University of Minnesota, Minneapolis, MN. Wilde, J. (2000). Identiﬁcation of multiple equation probit models with endogenous dummy regressors. Economics Letters, 69(3), 309–312. Zwick, T. (2002). Continuous training and ﬁrm productivity in Germany. ZEW Discussion Paper no. 02-50. Mannheim.

Effect of Multiskilling

61

APPENDIX. VARIABLE DEFINITIONS Labor Productivity: Dummy equaling 1 if the establishment is reported to have labor productivity above the industry average (¼0 if at or below the industry average). Product Quality: Dummy equaling 1 if the establishment is reported to produce products of a quality above the industry average (¼0 if at or below the industry average). Financial Performance: Dummy equaling 1 if the establishment is reported to have ﬁnancial performance above the industry average (¼0 if at or below the industry average). Multiskilling¼1 if at least some core production workers are multiskilled (¼0 otherwise). Volatility¼1 if the current state of the market for the main product or service of the establishment is described as ‘‘turbulent’’ (¼0 if described as ‘‘growing,’’ ‘‘mature,’’ or ‘‘declining’’). Establishment Size: Number of workers at the establishment. Union: Dummy equaling 1 if there are any workers at the establishment covered by a union (¼0 otherwise). Fraction of Part-Time Workers: Fraction of part-time workers at the establishment. Fraction of Temporary Workers: Fraction of temporary workers at the establishment. Percent Union: Percentage of workers at the establishment covered by a union. Number of Recognized Unions: (at the establishment) Owner Manager: Dummy equaling 1 if there is an owner-manager (¼0 otherwise). Foreign Owned: Dummy equaling 1 if foreign owned (¼0 otherwise). Establishment is At Least 5 Years Old: Dummy equaling 1 if this is true (¼0 otherwise). Franchise: Dummy equaling 1 if establishment is a franchise (¼0 otherwise).

62

MARTIN FARNHAM AND EMMA HUTCHINSON

Fixed-Term Percentage: Percentage of workers on ﬁxed-term contracts at the establishment. Fixed Term: Dummy equaling 1 if any workers are on ﬁxed-term contracts (¼0 otherwise). Temporary Workers: Dummy equaling 1 if any temporary workers at the establishment (¼0 otherwise). Industry Categories: (Manufacturing; Electricity, Gas, and Water; Construction; Wholesale and Retail; Hotels and Restaurants; Transport and Communication; Financial Services; Other Business Services; Public Administration; Education; Health; Other Community Services). Largest Occupational Group at the Establishment Categories: Professionals; Associate Professional and Technical; Administrative and Secretarial; Skilled Trades; Caring, leisure, and personal service; Sales and customer service; Process, Plant and Machine Operatives and Drivers; Routine Unskilled.

TEAMS, AUTONOMY, AND THE FINANCIAL PERFORMANCE OF FIRMS: NEW EVIDENCE FROM PANEL DATA Jed DeVaro ABSTRACT I use the 1998–2004 Workplace Employee Relations Survey (WERS) panel data set of British establishments to analyze the effect of team production (either autonomous/self-managed or closely managed) on ﬁnancial performance. The pattern of evidence is consistent with a positive association between team production in an establishment’s largest occupational group and the likelihood of improved ﬁnancial performance for that establishment. However, the results are mixed concerning whether the positive effects of teams are larger for autonomous teams versus nonautonomous teams. Keywords: Self-managed teams; ﬁnancial performance JEL classiﬁcations: M50; J0

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 63–85 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012007

63

64

JED DEVARO

INTRODUCTION In earlier work, I explored the effect of team production on organizational performance (DeVaro, 2006, 2008; DeVaro & Kurtulus, 2006), using crosssectional establishment-level data from the 1998 British Workplace Employee Relations Survey (WERS). In the ﬁrst of these studies (DeVaro, 2006), I analyzed the effect of team production on ﬁnancial performance and whether and how this effect varied by whether the teams were selfmanaged or closely managed. The main result of the paper was that higher ﬁnancial performance was indeed associated with the use of teams, although there was no evidence that self-managed (or autonomous) teams were more beneﬁcial than closely managed teams as is commonly argued in the vast empirical literature on the effects of teams on organizational performance. The paper differed from most of the prior literature in two ways. First, it estimated empirical models in which both teams and autonomy were treated as endogenous along with ﬁnancial performance. Second, its use of ﬁnancial performance as the dependent variable was unusual; the prior literature has been dominated by intermediate outcome measures (e.g., labor productivity) and ﬁnancial performance is rarely if ever seen. Financial performance (interpreted as synonymous with proﬁt in the WERS sample I used) is ultimately of interest in assessing whether teams, autonomous or otherwise, are beneﬁcial to organizations on average. This is because ﬁnancial performance incorporates the full spectrum of beneﬁts and costs associated with teams, in contrast to intermediate measures such as labor productivity. One limitation of my earlier studies is that they are based on only a single cross-section, so that all of the identifying information from the data derives from variation across employers in the nature and use of team production in a given year. The most recent wave of the WERS offers an opportunity to address this limitation, given that the 1998–2004 panel is the ﬁrst WERS panel containing information on teams, autonomy, and ﬁnancial performance. My goal in this chapter is to use the 1998–2004 WERS panel to reinvestigate the question of whether and how teams affect ﬁnancial performance and whether and how this effect varies by whether teams are granted autonomy. The analysis exploits temporal variation in the use of teams and autonomy within production units as well as variation across production units at a point in time. I focus only on ﬁnancial performance, as opposed to labor productivity and product quality that I studied in DeVaro (2008), because neither of these intermediate measures of organizational performance is included in the 1998–2004 panel. Consistent with my earlier work using the 1998 WERS cross-section, the pattern of results suggests that

Teams, Autonomy, and the Financial Performance of Firms

65

teams are associated with a greater likelihood of improved ﬁnancial performance over the period studied, but the evidence regarding whether autonomous or nonautonomous teams are better is mixed and inconclusive. The remainder of the chapter is organized as follows. In the second section, I provide some background on the previous literature and the theoretical channels through which teams – autonomous or closelymanaged – can be expected to inﬂuence organizational performance. In the third section, I describe the 1998–2004 WERS panel data set on which the analysis is based. In the fourth section, I present the empirical analysis, and in the ﬁfth section, I discuss some limitations of the analysis and then conclude with some suggestions for a program for collection of new data.

BACKGROUND AND PREVIOUS LITERATURE An economic approach to research on teams requires weighing the beneﬁts of teams to organizational performance against the costs. It is frequently argued, in both academic and managerial circles, that the beneﬁts of team production outweigh the costs more often than not, and that the case for teams is even stronger when they are autonomous or self-managed than when they are nonautonomous or closely managed. Teams, particularly selfmanaged ones, are frequently cited in textbook discussions of the practices for achieving high-commitment human resource management (Baron & Kreps, 1999, p. 89). A critique of the teams literature is that the costs of team production tend to receive less emphasis and attention than the beneﬁts, particularly outside of economics. At least part of the reason why teams, and particularly self-managed teams, are thought to be beneﬁcial derives from the dependent variables that are typically used in teams research. As discussed in Cohen and Bailey (1997), there are three broad sets of dependent variables that appear in the literature: attitudinal outcomes such as employee satisfaction, commitment, and trust in management; behavioral outcomes such as worker absenteeism, turnover, and safety; and organizational performance outcomes such as quantity and quality of outputs, efﬁciency, labor productivity, sales, response time, wages, customer satisfaction, innovation, and ﬁnancial performance. As argued in Cohen and Bailey, when the literature as a whole is reviewed, the estimated effects of teams are generally positive when the dependent variable is either an attitudinal measure or a behavioral measure, and the positive effects are stronger for self-managed teams more often than not. However, when the dependent variable is a measure of organizational

66

JED DEVARO

performance, the evidence is less clear. Therefore, the commonly held view that teams, particularly self-managed ones, are beneﬁcial likely arises from thinking of this literature as a whole, grouping all dependent variables, even though the picture is less clear when we focus only on organizational performance measures. All of this is to say that empirical analyses on the effects of teams, autonomous or otherwise, on organizational performance, are still of interest despite the vast literature on teams that already exists. The empirical literature on teams in economics focuses almost entirely on dependent variables that reﬂect organizational performance, particularly labor productivity and product quality. A summary of that literature can be found in DeVaro (2008). In some studies, teams are studied in isolation and in others as part of broader systems of high-performance work practices (Boning, Ichniowski, & Shaw 2003; DeVaro 2006, 2008; Encinosa, Gaynor, & Rebitzer, 2000; Gaynor & Gertler, 1995; Hamilton, Nickerson, & Owan 2003; Hansen, 1997; Ichniowski, Shaw, & Prennushi, 1997; Jones, Kalmi, & Kauhanen 2010; Jones & Kato, 2011; Leibowitz & Tollison, 1980; Liberti, 2003; Nalbantian & Schotter, 1997). Financial performance, as studied in this chapter in a panel context and in DeVaro (2008) in a cross-sectional context, is of particular interest as an outcome variable, given that it accounts for the full spectrum of beneﬁts and costs associated with team production. The potential beneﬁts of teams to organizational performance are well known, accruing through information sharing among workers and the complementarities arising when workers have knowledge that is nonduplicative and relevant to the production process (Lazear, 1995, 1998). Against the potential beneﬁts are costs, particularly those associated with free riding (Alchian & Demsetz, 1972; Holmstrom, 1982; Itoh, 1991, 1992; Legros & Matthews, 1993; McAfee & McMillan, 1991; Rasmusen, 1987). Granting authority, either to individual workers or to teams, also is associated with beneﬁts and costs that have been extensively explored in the theoretical literature in economics and other disciplines. One potential beneﬁt, as discussed in the organizational behavior literature, is enhanced worker motivation (Hackman, 1987). Another rationale for autonomy derives from socio-technical systems theory (e.g., Pearce & Ravlin, 1987). This literature argues that team members jointly optimize the social and technical systems of the organization and that the whole exceeds the sum of the parts when there is ‘‘ﬁt’’ between these systems. Another potential beneﬁt is discussed in Batt (2001), namely, the cost savings that can arise from eliminating supervisory roles. Whether it makes sense for a ﬁrm to

Teams, Autonomy, and the Financial Performance of Firms

67

grant teams autonomy should depend in large part on which means of monitoring workers is most efﬁcient for the production context. There is evidence that in some settings, the peer monitoring that occurs in selfmanaged teams is superior to supervisory monitoring by encouraging workers to meet the norms self-imposed by the group (Barker, 1993). The economics literature concerning delegation of authority within organizations is also relevant to a discussion of the beneﬁts and costs of granting autonomy to teams. Delegation of authority can be thought of as conferring ownership of an asset to which the owner has decision rights regarding its use (Grossman & Hart, 1986; Hart, 1995; Hart & Moore, 1990). An idea that emerges from this literature is the notion that agents (e.g., team members) have weaker incentives to take actions that will beneﬁt the ﬁrm if they are denied decision-making authority. Endowments of information, particularly asymmetric endowments, are key to such models. For example, Aghion and Tirole (1997) argue that a principal who has formal authority over a decision or activity can choose to reverse a subordinate’s decision but will avoid doing so if the subordinate is better informed and if their objectives are correctly aligned. Another incentivebased paper focusing on the impact of authority on the information structure of the organization is Stein (2002). Stein draws a distinction between hard and soft information. When information is hard (or veriﬁable), there is no downside to maintaining a hierarchical structure as opposed to delegating authority. In this model, it is the absence of authority that creates incentives for workers, because they are motivated to produce more positive information for the management en route to making a stronger case for a larger share of the capital budget. Although Aghion and Tirole (1997) and Stein (2002) focus on the impact of authority on the information structure, Dessein (2002) provides a purely informational rationale for delegation, taking the information structure (in which the agent is better informed than the principal) as given. The agent, if granted authority, will make decisions that make use of all the privately held information. A prediction that emerges is that centralization of authority is only optimal if the principal has the information relevant to the main decisions or is able to verify the information provided by lower levels of the hierarchy. The central trade-off in Dessein’s model is between a loss of control under delegation and a loss of information under communication. Thus, delegation of authority serves as an alternative to communication of information. The upshot of this discussion is that theoretical arguments can be made for and against the delegation of authority in a team context (or an individual worker context), and it is an

68

JED DEVARO

empirical question how these various forces impact ﬁnancial performance, on average, across organizations.

DATA AND MEASURES The data are from the management questionnaire in the 1998–2004 panel of the British WERS, jointly sponsored by the Department of Trade and Industry, ACAS, the Economic and Social Research Council, and the Policy Studies Institute. Distributed through the UK Data Archive, the WERS data are a nationally representative stratiﬁed random sample covering British workplaces with at least 10 employees except for those in the following 1992 Standard Industrial Classiﬁcation (SIC) divisions: agriculture, hunting, and forestry; ﬁshing; mining and quarrying; private households with employed persons; and extraterritorial organizations. Some of the 3,192 workplaces targeted were found to be out of scope, and the ﬁnal sample size of 2191 implies a net response rate of 80.4% (Cully, Woodland, O’Reilly, & Dix, 1999) after excluding the out-of-scope cases. Data were collected between October 1997 and June 1998 through face-to-face interviews, and the respondent manager was usually the most senior manager at the workplace with responsibility for employment relations.

Financial Performance The 1998 cross-section asks respondents to rate the current ﬁnancial performance of the workplace relative to that of others in the same industry. Responses include ‘‘A lot better than average,’’ ‘‘Better than average,’’ ‘‘About average for industry,’’ ‘‘Below average,’’ ‘‘A lot below average,’’ and ‘‘No comparison possible.’’ The ﬁnancial performance question asked in 2004 for the 1998–2004 panel is somewhat different, and therefore, the same variable is not observed in both years. Respondents in 2004 are asked whether ﬁnancial performance in the industry as a whole has improved, stayed the same, or deteriorated since 1998. Then, for each of those three possible responses, the respondent is asked about the improvement (or deterioration) of the workplace in question relative to the change in the industry. This information is insufﬁcient to create a 2004 ﬁnancial performance variable deﬁned in exactly the same way as the 1998 variable, and for that reason, an analysis that incorporates establishment-level individual effects is not possible. Instead, I deﬁne the following discrete

Teams, Autonomy, and the Financial Performance of Firms

69

dependent variable indicating whether the workplace’s 2004 ﬁnancial performance position (relative to the industry average) has improved, stayed the same, or deteriorated since 1998: Financial Performance=1 if (ﬁnancial performance for establishments in the same industry has improved since 1998 and, compared with that improvement, this workplace’s ﬁnancial performance has improved at a slower rate, or remained static, or actually deteriorated) or (ﬁnancial performance for establishments in the same industry has stayed the same since 1998 and, compared with that stability, this workplace’s ﬁnancial performance has deteriorated) or (ﬁnancial performance for establishments in the same industry has deteriorated since 1998 and, compared with that deterioration, this workplace’s ﬁnancial performance has deteriorated at a faster rate). = 2 if (ﬁnancial performance for establishments in the same industry has improved since 1998 and, compared with that improvement, this workplace’s ﬁnancial performance has improved at a similar rate) or (ﬁnancial performance for establishments in the same industry has stayed the same since 1998 and, compared with that stability, this workplace’s ﬁnancial performance has remained stable like the rest of the industry) or (ﬁnancial performance for establishments in the same industry has deteriorated since 1998 and, compared with that deterioration, this workplace’s ﬁnancial performance has deteriorated at the same rate as the rest of the industry). = 3 if (ﬁnancial performance for establishments in the same industry has improved since 1998 and, compared with that improvement, this workplace’s ﬁnancial performance has improved at a faster rate) or (ﬁnancial performance for establishments in the same industry has stayed the same since 1998 and, compared with that stability, this workplace’s ﬁnancial performance has improved) or (ﬁnancial performance for establishments in the same industry has deteriorated since 1998 and, compared with that deterioration, this workplace’s ﬁnancial performance has remained stable or actually improved since 1998). Teams Starting with the 1998 cross-section, the WERS asked questions about the nature and use of teams. The same questions are asked in the 2004, both for the cross-section and for the 1998–2004 panel. The respondent manager is asked to report the proportion of employees in the largest occupational group at the workplace that works in formally designated teams. Responses are in the following discrete categories: ‘‘All 100%,’’ ‘‘Almost all 80–99%,’’

70

JED DEVARO

‘‘Most 60–79%,’’ ‘‘Around half 40–59%,’’ ‘‘Some 20–39%,’’ ‘‘Just a few 1–19%,’’ and ‘‘None 0%.’’ Following my earlier studies of teams, in both years, I aggregate this variable to a binary indicator of whether teams are used in the largest occupational group. I then deﬁne the following four dummies: Teams00 =1 if teams are not used in the largest occupational group in either 1998 or 2004 =0 otherwise Teams01 =1 if teams are used in the largest occupational group in 2004 but not in 1998 =0 otherwise Teams10 =1 if teams are used in the largest occupational group in 1998 but not in 2004 =0 otherwise Teams11 =1 if teams are used in the largest occupational group in 1998 and 2004 =0 otherwise There are two reasons for aggregating the original seven-valued teams measure to a binary variable. First, classiﬁcation errors can be expected to be smaller for the binary measure (that simply captures whether teams are used, or not, in the largest occupational group) than for the original measure. This is particularly important given that the teams questions pertain to the establishment’s largest occupational group. For example, consider two establishments, A and B, each of the same size. In A, the establishment’s largest occupational group comprises two-thirds of the establishment’s workforce, whereas in B, it comprises less than 10%. If A responds that ‘‘Some (20%–39%)’’ of its largest occupational group is in teams, whereas B responds that ‘‘Around half (40%–59%) of its largest occupational group is in teams, a seven-valued teams variable would indicate that teams are more prevalent in B, whereas the reverse is true at the establishment level (and note that the dependent variable pertains to organizational performance at the establishment level). This type of classiﬁcation problem is mitigated by considering a binary team measure. A second reason for aggregating the teams variable to two categories is parsimony. Standard errors are large in most speciﬁcations even using binary measures of teams and autonomy, and therefore, more general models are impractical.

Teams, Autonomy, and the Financial Performance of Firms

71

Autonomy The measure of autonomy is given by the following WERS question: For establishments that report the use of formally designated teams in the largest occupational group, the respondent manager is asked to respond ‘‘Yes’’ or ‘‘No’’ to the following statement: ‘‘Team members jointly decide how the work is to be done.’’ This was also the primary autonomy measure I investigated in DeVaro (2006), and it corresponds closely to the notion of autonomy used in the interdisciplinary literature on self-managed teams. It is argued in the organizational behavior literature that team members are motivated when ‘‘the task provides group members with substantial autonomy for deciding about how they do the work [emphasis added] – in effect, the group ‘owns’ the task and is responsible for the work outcomes’’ (Hackman, 1987, p. 324). In the economics literature, notions of autonomy, at the level of both individual workers and teams, are similar, focusing on the right of workers to select actions (i.e., tasks) affecting part or the whole of an organization (Simon, 1951). From the autonomy and the teams questions, I construct the following binary indicators: Teams01a

Teams01n

Teams1a0

Teams1n0

Teams1a1a Teams1n1n Teams1a1n

Teams1n1a

=1 if autonomous teams are used in the largest occupational group in 2004 but no teams are used in 1998 =0 otherwise =1 if nonautonomous teams are used in the largest occupational group in 2004 but no teams are used in 1998 =0 otherwise =1 if autonomous teams are used in the largest occupational group in 1998 but no teams are used in 2004 =0 otherwise =1 if nonautonomous teams are used in the largest occupational group in 1998 but no teams are used in 2004 =0 otherwise =1 if autonomous teams are used in both 1998 and 2004 =0 otherwise =1 if nonautonomous teams are used in both 1998 and 2004 =0 otherwise =1 if autonomous teams are used in 1998 and nonautonomous teams in 2004 =0 otherwise =1 if nonautonomous teams are used in 1998 and autonomous teams in 2004 =0 otherwise

72

JED DEVARO

Employer Characteristics Some of the empirical speciﬁcations of the next section include controls for employer characteristics, measured in 2004. These are as follows: the natural logarithm of establishment size (all workers currently on payroll at the establishment), a binary indicator for whether any of the employees at the establishment are unionized, an indicator for whether the establishment is a single-establishment ﬁrm, industry dummies (manufacturing; electricity, gas, and water; construction; wholesale and retail; hotels and restaurants; transport and communication; ﬁnancial services; other business services; public administration; education; health; and other community services), and dummies for the establishment’s largest occupational group (managers and administrators, professional occupations, associate professional and technical occupations, clerical and secretarial occupations, craft and related occupations, personal and protective service occupations, sales occupations, plant and machine operatives, and other occupations).

EMPIRICAL TESTS The ﬁrst part of the analysis aims to measure the effect of teams on ﬁnancial performance, without regard to whether the teams are autonomous or nonautonomous. I estimate an ordered probit model in which Financial Performance is the dependent variable, and the key independent variables are Teams00, Teams01, and Teams10. I report the results of four speciﬁcations. The ﬁrst includes only the aforementioned independent variables. The second includes those independent variables plus controls for ﬁrm characteristics. The third and fourth repeat the ﬁrst and second, respectively, using a more restricted subsample that requires that each establishment’s largest occupational group be the same in 2004 as it was in 1998. This is potentially important. For example, suppose the observed teams (or autonomy) variable changes in 2004 from its value in 1998. This could be because of a true change or because the various occupations at the establishment changed in their relative sizes and that the new (i.e., 2004) largest occupational group is different from the old (i.e., 1998) one with respect to its use of teams or autonomy. In this case, it would misleadingly appear that the employer was instituting changes in teams or autonomy. The sample restriction eliminates this concern, at the expense of sacriﬁcing over 30% of the observations in models where precision is already low for some of the key parameters. For this reason, I present both sets of results. Descriptive statistics are presented in Table 1, and results for the four speciﬁcations are displayed in Table 2. The theoretically predicted effect of

73

Teams, Autonomy, and the Financial Performance of Firms

Table 1.

Descriptive Statistics. Fractions

Financial performance =1 =2 =3 Teams and autonomy Teams00 Teams01 Teams10 Teams11 Teams01a Teams01n Teams1a0 Teams1n0 Teams1a1a Teams1n1n Teams1a1n Teams1n1a Employer characteristics Establishment size Union Single-establishment ﬁrm Industry Manufacturing Electricity, gas, and water Construction Wholesale and retail Hotels and restaurants Transport and communication Financial services Other business services Public administration Education Health Other community services Largest occupational group at workplace Executives Professionals Associate professional and technical Administrative and secretarial occupations Skilled trades Caring, leisure, and personal service Sales and customer service

Mean

Standard Error

0.140 0.114 0.109 0.637 0.069 0.065 0.090 0.035 0.322 0.094 0.174 0.149

0.029 0.023 0.023 0.022 0.035 0.022 0.024 0.010 0.033 0.014 0.031 0.027

74.469 0.536 0.297

4.728 0.035 0.034

0.129 0.002 0.040 0.130 0.089 0.051 0.041 0.108 0.064 0.145 0.152 0.049

0.023 0.001 0.012 0.027 0.022 0.017 0.014 0.023 0.016 0.022 0.024 0.012

0.012 0.156 0.061 0.190 0.085 0.230 0.110

0.006 0.023 0.013 0.029 0.017 0.029 0.026

0.108 0.493 0.400

74

JED DEVARO

Table 1. (Continued ) Fractions Process, plant and machine operatives and drivers Routine unskilled

Mean

Standard Error

0.107 0.051

0.024 0.014

Sample size=935 Notes: Statistics are computed on the full sample. Due to missing values, the sample size varies across variables; all non-missing observations were used to compute each statistic. Summary statistics based on the analysis subsample are available upon request and are similar to those reported here.

Table 2.

Effect of Teams on Financial Performance (Ordered Probit Coefﬁcients from Noninteractive Speciﬁcations). Dependent Variable: Financial Performance

Teams00 Teams01 Teams10

Main sample

Main sample

Restricted sample

Restricted sample

0.066 (0.229) 0.204 (0.348) 0.417 (0.280)

0.098 (0.239) 0.156 (0.274) 0.448 (0.259)

0.220 (0.271) 0.058 (0.424) 0.432 (0.329)

0.180 (0.276) 0.005 (0.311) 0.460 (0.302)

No No

0.210 (0.068) 0.413 (0.160) 0.151 (0.162) Yes Yes

No No

0.362 (0.079) 0.344 (0.186) 0.107 (0.189) Yes Yes

858

857

593

593

Employer characteristics ln(Establishment size) Union Single-establishment ﬁrm Industry controls Largest occup. group controls Sample size

Note: Cell entries are ordered probit coefﬁcients with standard errors in parentheses.

teams on ﬁnancial performance, given the previous literature, is positive. Although most of the prior empirical literature ﬁnding support for this positive predicted effect focuses on intermediate measures of organizational performance (e.g., labor productivity), in my earlier study using the 1998

Teams, Autonomy, and the Financial Performance of Firms

75

WERS cross-section (DeVaro, 2006), I found a positive effect on ﬁnancial performance. This directional hypothesis suggests one-tailed signiﬁcance tests, which I use throughout the analysis. Across each of the four speciﬁcations, only the coefﬁcient of Teams10, which is negative, is signiﬁcant at the 10% level. The result suggests that, relative to establishments that used teams in the largest occupational group both in 1998 and in 2004, those that started with teams in 1998 and then abandoned them between 1998 and 2004 had a lower probability of experiencing improvements in ﬁnancial performance relative to the industry average during 1998–2004. The result is consistent with the hypothesized positive effect of team production on ﬁnancial performance. Note that the magnitude of the Teams10 coefﬁcient is largely unchanged across speciﬁcations. This insensitivity of the coefﬁcient to the speciﬁcation might be used as the basis for focusing on the models that exploit the larger sample size without the sample restriction (i.e., speciﬁcations 1 and 2, as opposed to 3 and 4). To be conservative, however, my preference is to focus on speciﬁcation 4 when drawing inferences from Table 2 and from subsequent analyses. I also investigate whether the effect of teams on ﬁnancial performance is affected by the starting level (i.e., the 1998 level) of ﬁnancial performance. I do this by ﬁrst constructing two binary variables; the ﬁrst equals 1 if 1998 ﬁnancial performance was above the industry average (and 0 otherwise), and the second equals 1 if 1998 ﬁnancial performance was ‘‘a lot above’’ the industry average (and 0 otherwise). I interact these two dummies with each of Teams00, Teams01, and Teams10, producing six two-way interactions that I add to each of the four empirical speciﬁcations discussed in Table 2. Results are displayed in Table 3. Three results emerge from the table. First, focusing on speciﬁcation 4, the earlier result from Table 2 regarding Teams10 appears to vanish for establishments whose 1998 ﬁnancial performance was above the industry average, so that the lower probabilities of improved ﬁnancial performance between 1998 and 2004 that are experienced by establishments that start out with teams and subsequently abandon them apply only to those establishments that started the period with ﬁnancial performance at or below the industry average. Although this result qualiﬁes and reﬁnes the earlier result from Table 2, in that it does not apply to establishments whose 1998 ﬁnancial performance was above the industry average, it is still consistent with the expectation from the prior literature of a positive effect of teams. Second, in contrast to the earlier result from Table 2, the coefﬁcient of Teams01 in speciﬁcation 4 is positive and statistically signiﬁcant, suggesting that employers that did not use teams in the largest occupational group in

76

JED DEVARO

Table 3.

Effect of Teams on Financial Performance (Ordered Probit Coefﬁcients from Interactive Speciﬁcations). Dependent Variable: Financial Performance

Teams00 Teams01 Teams10 Teams00 Finper98(4) Teams01 Finper98(4) Teams10 Finper98(4) Teams00 Finper98(5) Teams01 Finper98(5) Teams10 Finper98(5) Finper98(4) Finper98(5)

Main sample

Main sample

Restricted sample

Restricted sample

0.061 (0.383) 0.424 (0.473) 0.697 (0.504) 0.266 (0.571) 0.851 (0.684) 0.824 (0.599) 0.139 (0.676) 0.092 (0.695) 0.703 (0.861) 0.255 (0.184) 0.166 (0.355)

0.185 (0.387) 0.465 (0.436) 0.627 (0.454) 0.381 (0.614) 0.810 (0.532) 0.621 (0.589) 0.170 (0.598) 0.293 (0.655) 0.769 (0.678) 0.286 (0.168) 0.050 (0.294)

0.636 (0.506) 1.029 (0.451) 0.996 (0.579) 0.194 (0.670) 1.361 (0.730) 1.284 (0.711) 1.355 (0.555) 0.960 (0.714) 0.358 (0.890) 0.328 (0.211) 0.103 (0.283)

0.775 (0.423) 0.836 (0.466) 0.847 (0.523) 0.522 (0.626) 0.890 (0.585) 1.008 (0.722) 1.325 (0.550) 1.093 (0.693) 0.522 (0.717) 0.360 (0.189) 0.055 (0.284)

No No

0.210 (0.064) 0.419 (0.174) 0.051 (0.166) Yes Yes

No No

0.303 (0.074) 0.275 (0.217) 0.001 (0.197) Yes Yes

735

734

509

509

Employer characteristics ln(Establishment size) Union Single-establishment ﬁrm Industry controls Largest occup. group controls Sample size

Notes: Cell entries are ordered probit coefﬁcients with standard errors in parentheses. Finper98(4) is a binary variable equaling 1 if the establishment’s 1998 ﬁnancial performance is above the industry average, and 0 otherwise. Finper98(5) is a binary variable equaling 1 if the establishment’s 1998 ﬁnancial performance is ‘‘a lot above’’ the industry average, and 0 otherwise.

Teams, Autonomy, and the Financial Performance of Firms

77

1998 but added them prior to 2004 experienced, on average, a higher probability of improved ﬁnancial performance relative to employers that used teams consistently in both years. However, this result appears to be absent for those establishment that started the period with ﬁnancial performance above (or a lot above) the industry average, pertaining only to those establishments starting the period with ﬁnancial performance at or below the industry average. Overall, the result is consistent with a positive effect of teams, given that (if teams are indeed beneﬁcial) adding teams where they are absent can be expected to improve ﬁnancial performance more than continuing to use them where they are already present. Third, in contrast to the earlier result from Table 2, the coefﬁcient of Teams00 in speciﬁcation 4 of Table 3 is positive and statistically signiﬁcant, suggesting that employers that used teams neither in 1998 nor in 2004 in their establishment’s largest occupational group experienced, on average, a higher probability of improved ﬁnancial performance relative to employers that used teams consistently in both years. This result appears counter to the hypothesis of a positive effect of teams on ﬁnancial performance. However, it should be noted that the result is clearly absent (and may even be reversed) for establishments that started the period with ﬁnancial performance a lot above the industry average. Furthermore, while I focus throughout on speciﬁcation 4 given that I see it as most conservative, it should be noted that the coefﬁcient of Teams00 is statistically insigniﬁcant in the other three speciﬁcations. Taken as a whole, the results from Table 3 seem consistent with a positive effect of teams on ﬁnancial performance, although there is less consistency across speciﬁcations of Table 3 than there was for Table 2. The second part of the analysis aims to distinguish between the effect of selfmanaged (or autonomous) teams and closely managed teams on ﬁnancial performance. I start by estimating noninteractive speciﬁcations that parallel those in Table 2, with the difference being that instead of including the three independent variables Teams00, Teams01, and Teams10, I use the larger set of indicators (distinguishing teams by whether or not they are granted autonomy) described in the previous section. The previous literature generally predicts that the beneﬁts of teams to organizational performance tend to be most pronounced when the teams are autonomous. An exception can be found in my earlier work (DeVaro, 2006) where there was a positive relationship between teams and ﬁnancial performance but no incremental positive effect of autonomy; this result was contrary to the conventional wisdom. However, it must be emphasized that most of the empirical teams literature has not used ﬁnancial performance as a dependent variable, but rather some intermediate measure such as labor productivity.

78

JED DEVARO

Results from the noninteractive speciﬁcations are displayed in Table 4. Again, in drawing inferences, I focus on speciﬁcation 4. The ﬁrst result echoes the earlier result from Table 2, namely that establishments that start the period with teams and subsequently abandon them experience a lower probability of improved ﬁnancial performance over the period than those that continue using teams. However, this result appears to be true only for autonomous teams. There is no evidence of a negative effect from abandoning teams if the teams were initially nonautonomous. This result is

Table 4. Effect of Autonomous versus Nonautonomous Teams on Financial Performance (Ordered Probit Coefﬁcients from Noninteractive Speciﬁcations). Dependent Variable: Financial Performance

Teams01a Teams01n Teams10 Teams1n0 Teams1n1n Teams1a1n Teams1n1a

Main sample

Main sample

Restricted sample

Restricted sample

0.099 (0.379) 0.582 (0.566) 0.842 (0.371) 0.045 (0.309) 0.455 (0.226) 0.161 (0.208) 0.257 (0.263)

0.171 (0.341) 0.587 (0.468) 0.910 (0.356) 0.248 (0.390) 0.625 (0.222) 0.265 (0.168) 0.343 (0.246)

0.012 (0.489) 0.358 (0.657) 0.906 (0.464) 0.082 (0.325) 0.352 (0.289) 0.050 (0.242) 0.421 (0.181)

0.038 (0.414) 0.263 (0.494) 0.912 (0.427) 0.139 (0.405) 0.551 (0.262) 0.090 (0.210) 0.492 (0.207)

No No

0.200 (0.072) 0.402 (0.168) 0.139 (0.173) Yes Yes

No No

0.345 (0.086) 0.340 (0.196) 0.0006 (0.184) Yes Yes

804

803

552

552

Employer characteristics ln(Establishment size) Union Single-establishment ﬁrm Industry controls Largest occup. group controls Sample size

Note: Cell entries are ordered probit coefﬁcients with standard errors in parentheses.

Teams, Autonomy, and the Financial Performance of Firms

79

consistent with the bulk of the literature arguing for the merits of autonomous teams. The second result is that employers that use nonautonomous teams in both years have a lower probability of improved ﬁnancial performance over the period than those using autonomous teams both periods. This result also favors granting teams autonomy, given that teams are used. However, the results also suggest that employers that use nonautonomous teams in 1998 and then switch to autonomous teams during 1998–2004 experience a lower probability of improved ﬁnancial performance, relative to the reference group. This result is more favorable to nonautonomous or closely managed teams. I also investigated whether the preceding effects depend on the level of ﬁnancial performance at the start of the period. To keep the number of twoway interactions manageable, I created only one dummy (as opposed to two in the analyses of Table 3) indicating whether current ﬁnancial performance in 1998 is above average for the industry or a lot above average for the industry (=1) versus at or below the industry average (=0). I interacted this dummy with the independent variables of interest from Table 4, reestimating all four speciﬁcations including all the interactions. Results are displayed in Table 5. The ﬁrst result from the table is that employers that use no teams in 1998 and subsequently introduce nonautonomous teams during 1998–2004 experience a higher probability of improved ﬁnancial performance during the period than employers who use autonomous teams in both years. However, this result only holds for employers with ﬁnancial performance at or below the industry average in 1998 and is reversed for higher-performing employers. It is interesting to note that the corresponding result does not hold when autonomous teams are introduced (given that no teams were used in 1998); in that case, there is no evidence of a difference relative to the reference group. The second result, which parallels a result from Table 4, is that employers that start with autonomous teams and subsequently abandon them in favor of no teams are less likely than the reference group to experience improved ﬁnancial performance during the period. However, this effect is mitigated for establishments with higher-than-average (relative to the industry) ﬁnancial performance in 1998. The third result, also paralleling a result from Table 4, is that employers using nonautonomous teams in both years have a lower probability of improved ﬁnancial performance during the period than the reference group. The fourth result, also paralleling a result from Table 4, is that employers that start off with nonautonomous teams and then grant teams autonomy

80

JED DEVARO

Table 5. Effect of Autonomous versus Nonautonomous Teams on Financial Performance (Ordered Probit Coefﬁcients from Interactive Speciﬁcations). Dependent Variable: Financial Performance

Teams01a Teams01n Teams1a0 Teams1n0 Teams1n1n Teams1a1n Teams1n1a Teams01a Finper98(high) Teams01n Finper98(high) Teams1a0 Finper98(high) Teams1n0 Finper98(high) Teams1n1n Finper98(high) Teams1a1n Finper98(high) Teams1n1a Finper98(high) Finper98(high)

Main sample

Main sample

Restricted sample

Restricted sample

0.353 (0.527) 0.488 (0.669) 1.488 (0.647) 0.146 (0.478) 0.596 (0.286) 0.160 (0.376) 0.800 (0.277) 0.807 (0.740) 1.933 (0.855) 0.840 (0.799) 0.170 (0.644) 0.188 (0.467) 0.083 (0.463) 0.670 (0.517) 0.251 (0.286)

0.485 (0.542) 0.517 (0.588) 1.410 (0.595) 0.213 (0.603) 0.741 (0.327) 0.302 (0.284) 0.780 (0.321) 0.911 (0.626) 1.973 (0.680) 0.609 (0.728) 0.177 (0.766) 0.101 (0.451) 0.035 (0.380) 0.605 (0.475) 0.337 (0.287)

0.216 (0.583) 1.144 (0.591) 2.147 (0.698) 0.217 (0.520) 0.677 (0.342) 0.070 (0.457) 0.900 (0.340) 0.430 (0.850) 2.333 (0.824) 1.654 (0.895) 0.492 (0.685) 0.543 (0.570) 0.166 (0.510) 0.935 (0.438) 0.449 (0.309)

0.169 (0.569) 1.044 (0.635) 1.951 (0.653) 0.192 (0.673) 0.933 (0.361) 0.003 (0.385) 0.760 (0.417) 1.013 (0.655) 2.004 (0.696) 1.538 (0.839) 0.286 (0.807) 0.687 (0.507) 0.074 (0.460) 0.766 (0.515) 0.552 (0.321)

No

0.215 (0.066) 0.388 (0.180) 0.051 (0.173) Yes

No

0.332 (0.078) 0.325 (0.217) 0.025 (0.202) Yes

Employer characteristics ln(Establishment size) Union Single-establishment ﬁrm Industry controls

81

Teams, Autonomy, and the Financial Performance of Firms

Table 5. (Continued ) Dependent Variable: Financial Performance

Largest occup. group controls Sample size

Main sample

Main sample

Restricted sample

Restricted sample

No 690

Yes 689

No 476

Yes 476

Notes: Cell entries are ordered probit coefﬁcients with standard errors in parentheses. Finper98(high) is a binary indicator equaling 1 if the establishment’s 1998 ﬁnancial performance is above the industry average, and 0 otherwise.

during the period experience a lower probability of improved ﬁnancial performance during the period, relative to the employers in the reference group. However, this is true only for establishments with ﬁnancial performance that is at or below the industry average in 1998. For such employers, the result is more favorable to nonautonomous or closely managed teams than to autonomous or self-managed ones.

DISCUSSION AND CONCLUSION The results of the previous section are of interest because, in contrast to my earlier work on the same subject using only cross-sectional data, the present results exploit variation not just across production units but also over time. Aiding comparability to my earlier results is the fact that the setting is the same (a random sample of British establishments, ﬁrst sampled in the year of my original studies) and that most of the key measures are the same. In particular, I focus on ﬁnancial performance, which is a rare outcome variable in an empirical literature that focuses more heavily on intermediate measures of organizational performance. Another positive feature of the present panel is that while it consists of only two years, they are six years apart. Although temporal variation in HR practices is always desirable in addition to crosssectional variation, there is usually little within-establishment variation in practices from one year to the next. But six years is a long enough span to generate a decent amount of temporal variation in the practices of interest. Despite these advantages, the results of the previous section must be qualiﬁed on several grounds. First and foremost, most of the results are estimated with low precision. In part, this is a consequence of the much

82

JED DEVARO

smaller sample sizes associated with the 1998–2004 WERS as opposed to the WERS cross-sections. Another factor likely contributing to low precision is the fact that the key questions about practices pertain not the establishment as a whole but rather to the establishment’s largest occupational group, whereas the dependent variable is measured at the establishment level. Yet another factor potentially contributing to low precision is the dependent variable itself; while it is interesting to examine as broad an outcome variable as ﬁnancial performance, it might be expected that effects of teams (and autonomy) might be harder to detect in this case than for an intermediate measure of performance (e.g., labor productivity) that is more directly impacted by HR practices and subject to a smaller set of stochastic inﬂuences. Second, although temporal variation in the practices of interests could be exploited in the 1998–2004 panel, I was unable to incorporate establishment-level individual effects into any of the statistical models. The reason for this is that the dependent variable measuring ﬁnancial performance that was measured in 1998 was not measured in the same way in 2004. Third, to be conservative, I have focused on the most controlled speciﬁcations that restrict the sample to those establishments for which the largest occupational group is the same in both 1998 and 2004. But it should be noted that the results sometimes differ across the other speciﬁcations, which is another consequence of low precision. Although I have focused on the fourth speciﬁcation in assessing the overall pattern of results, other readers of the same results may have a different interpretation that assigns different relative weights to the four speciﬁcations. Low precision of the estimates notwithstanding, some patterns in the data emerge that are worth noting in light of the previous literature. In particular, the results seem consistent with a positive effect of teams on ﬁnancial performance, which is consistent with the ﬁndings from my cross-sectional work on this question in DeVaro (2006). The second result, also consistent with my earlier cross-sectional work in DeVaro (2006), is that results are mixed concerning whether autonomous or nonautonomous teams are better in terms of ﬁnancial performance. Some results are more favorable to autonomous teams and some to nonautonomous teams. These results suggest that granting teams autonomy is less essential than commonly argued in the literature in terms of beneﬁts to ﬁnancial performance, although it must again be noted that ﬁnancial performance rarely appears in this literature as an outcome measure. The 1998–2004 WERS panel offers a rare opportunity to analyze these questions in a panel setting; usually such panels involve much narrower samples involving one or a small number of ﬁrms. Unfortunately, low

Teams, Autonomy, and the Financial Performance of Firms

83

precision is a signiﬁcant impediment to extracting robust insights from this sample. Looking to the future, it is hoped that future waves of WERS and similar data sets build on the strengths of the 1998–2004 panel while repairing some of the weak points. In particular, it would be helpful for outcome variables such as ﬁnancial performance to be measured in the same way in each year so that individual effects can be incorporated into statistical models. It would also be helpful to step away from the focus on the establishment’s ‘‘largest occupational group,’’ given that this creates some incompatibility between the outcome measure (which is at the establishment level) and the key independent variables; this would also eliminate the problems that arise when the largest occupational group changes across years in the panel. Finally, it would be helpful to maintain the usual sample sizes but restricting the sampling frame to a particular industry. This would have the advantage of increasing precision and sharpening results by cutting down on the signiﬁcant heterogeneity that exists across production units in a nationally representative random sample, while at the same time avoiding the negatives of the other extreme, which is convenience samples of one or a small number of ﬁrms that may yield highly idiosyncratic results. My hope is that this chapter will stimulate collection of such samples in the future.

ACKNOWLEDGMENTS The author acknowledges the Department of Trade and Industry, the Economic and Social Research Council, the Advisory, Conciliation and Arbitration Service and the Policy Studies Institute as the originators of the 2004 Workplace Employee Relations Survey data, and the Data Archive at the University of Essex as the distributor of the data. None of these organizations bears any responsibility for the author’s analysis and interpretations of the data. I gratefully acknowledge helpful feedback from Takao Kato.

REFERENCES Aghion, P., & Tirole, J. (1997). Formal and real authority in organizations. Journal of Political Economy, 105(February), 1–29. Alchian, A. A., & Demsetz, H. (1972). Production, information costs, and economic organization. American Economic Review, 62(December), 777–795.

84

JED DEVARO

Barker, J. R. (1993). Tightening the iron cage: Concertive control in self-managing teams. Administrative Science Quarterly, 38, 408–437. Baron, J. N., & Kreps, D. M. (1999). Strategic human resources: Frameworks for general managers. New York, NY: Wiley. Batt, R. (2001). The economics of teams among technicians. British Journal of Industrial Relations, 39(March), 1–24. Boning, B., Ichniowski, C., & Shaw, K. (2003). Opportunity counts: Teams and the Effectiveness of production incentives. Working Paper no. 8306. National Bureau of Economic Research, Cambridge, MA. Cohen, S., & Bailey, D. (1997). What makes teams work: Group effectiveness research from the shop ﬂoor to the executive suite. Journal of Management, 23(May–June), 239–290. Cully, M., Woodland, S., O’Reilly, A., & Dix, G. (1999). Britain at work: As depicted by the 1998 workplace employee relations survey. London: Routledge. Dessein, W. (2002). Authority and communication in organizations. The Review of Economic Studies, 69(October), 811–838. DeVaro, J. (2006). Teams, autonomy, and the ﬁnancial performance of ﬁrms. Industrial Relations: A Journal of Economy & Society, 45(2), 217–269. DeVaro, J. (2008). The effect of self-managed and closely-managed teams on labor productivity and product quality: An empirical analysis of a cross section of establishments. Industrial Relations: A Journal of Economy & Society, 47(4), 659–697. DeVaro, J., & Kurtulus, F. A. (2006). What types of organizations beneﬁt from team production, and how do they beneﬁt? Advances in the Economic Analysis of Participatory and Labor-Managed Firms, 9, 3–56. Encinosa, W.E., Gaynor, M., & Rebitzer, J. B. (2000). The sociology of groups and the economics of incentives: Theory and evidence on compensation systems. Mimeo. Carnegie Mellon University, Pittsburgh, PA. Gaynor, M., & Gertler, P. J. (1995). Moral hazard and risk spreading in medical partnerships. RAND Journal of Economics, 26(Winter), 591–613. Grossman, S. J., & Hart, O. (1986). The costs and beneﬁts of ownership: A theory of vertical and lateral integration. Journal of Political Economy, 94(August), 691–719. Hackman, J. R. (1987). The design of work teams. In: J. Lorsch (Ed.), Handbook of organizational behavior. Englewood Cliffs, NJ: Prentice-Hall. Hamilton, B. H., Nickerson, J. A., & Owan,, H. (2003). Team incentives and worker heterogeneity: An empirical analysis of the impact of teams on productivity and participation. Journal of Political Economy, 111(June), 465–497. Hansen, D. (1997). Worker performance and group incentives: A case study. Industrial & Labor Relations Review, 51(October), 37–49. Hart, O. (1995). Corporate governance: Some theory and implications. The Economic Journal, 105(May), 678–689. Hart, O., & Moore, J. (1990). Property rights and the nature of the ﬁrm. Journal of Political Economy, 98(December), 1119–1158. Holmstrom, B. (1982). Moral hazard in teams. Bell Journal of Economics, 13(Autumn), 324–340. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The effects of human resource management practices on productivity: A study of steel ﬁnishing lines. American Economic Review, 87(June), 291–313. Itoh, H. (1991). Incentives to help in multi-agent situations. Econometrica, 59(May), 611–636.

Teams, Autonomy, and the Financial Performance of Firms

85

Itoh, H. (1992). Cooperation in hierarchical organizations: An incentive perspective. Journal of Law, Economics, & Organization, 8(April), 321–345. Jones, D., & Kato, T. (2011). The impact of teams on output, quality and downtime: An empirical analysis using individual panel data. Industrial & Labor Relations Review, 64, 1–29. Jones, D. C., Kalmi, P., & Kauhanen, A. (2010). Teams, performance-related pay, proﬁtsharing and productive efﬁciency: Evidence from a food-processing plant. Industrial & Labor Relations Review, 63(4), 606–626. Lazear, E. P. (1995). Personnel economics. Cambridge, MA; London: MIT Press. Lazear, E. P. (1998). Personnel economics for managers. New York, NY: Wiley. Legros, P., & Matthews, S. A. (1993). Efﬁcient and nearly efﬁcient partnerships. The Review of Economic Studies, 60(July), 599–611. Leibowitz, A., & Tollison, R. (1980). Free riding, shirking and team production in legal partnerships. Economic Inquiry, 18(July), 380–394. McAfee, R. P., & John, M. (1991). Optimal contracts for teams. International Economic Review, 32(August), 561–577. Liberti, J. M. (2003). Initiative, incentives, and soft information. How does delegation impact the role of bank relationship managers? Ph.D. dissertation, University of Chicago, Chicago, IL. Nalbantian, H. R., & Schotter, A. (1997). Productivity under group incentives: An experimental study. American Economic Review, 87(June), 314–341. Pearce, J. A., & Ravlin, E. C. (1987). The design and activation of self-regulating work groups. Human Relations, 40, 751–781. Rasmusen, E. (1987). Moral hazard in risk-averse teams. RAND Journal of Economics, 18(Autumn), 428–435. Simon, H. A. (1951). A formal theory of the employment relationship. Econometrica, 19(July), 293–305. Stein, J. (2002). Information production and capital allocation: Decentralized versus hierarchical ﬁrms. The Journal of Finance, 57(October), 1891–1921. Workplace Employee Relations Survey: Cross-Section. (1998). [computer ﬁle]. 6th Edition. Department of Trade and Industry and Advisory, Conciliation and Arbitration Service. Colchester, Essex: UK Data Archive [distributor], January 2001. SN: 395.

PART II COMPENSATION, WORKER ATTITUDES, AND PRODUCTIVITY

HOW DO RULES AND COSTS AFFECT A FIRM’S SETTING OF BENEFITS? THE CASE OF HEALTH INSURANCE AND WORKFORCE SKILLS Nan L. Maxwell ABSTRACT Institutional rules and economies of scale can create incentives for ﬁrms to make inframarginal decisions when offering fringe beneﬁts. We examine how such incentives might affect a ﬁrm’s offer of health insurance. We develop and estimate an empirical model of the ﬁrm’s offer of health insurance that includes incentives created by rules and economies of scale. We quantify the behavioral manifestations from rules and costs as recruiting difﬁculty in areas outside those in which compensation is set and the percentage of high-skilled jobs in the ﬁrm and use the California Health and Employment Surveys (CHES) to estimate the model. We show a 10–13 percentage point increase in the probability of a ﬁrm offering workers health insurance in jobs outside of those in which compensation is being set, if the recruiting difﬁculty lies in mid- or highskilled positions. This increase is about twice the size of the increase

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 89–114 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012008

89

90

NAN L. MAXWELL

associated with recruiting difﬁculty in the position in which compensation is negotiated. A failure to control for the inﬂuence of inframarginal decision making when estimating the wage-insurance tradeoff helps produce wrong-signed estimates. By bringing institutional rules and economies of scale into the framework of a ﬁrm’s offer of fringe beneﬁts, we help move the focus of the fringe beneﬁt-wage tradeoff away from the individual level. Keywords: Fringe beneﬁts; employer–sponsored health insurance; wage-beneﬁt tradeoff; workforce skills JEL classiﬁcation: M52 When a worker shops for a job, two questions often are paramount: what wages will I receive and what beneﬁts will I be offered? Some workers, such as youth and those with a relatively low income may focus heavily on wages when weighing a job offer, whereas others such as heads of families and those with a relatively high income, may actively scrutinize the beneﬁt package offered. Because wages represented only 70.8 percent of worker compensation in 2009 (EBRI, 2010), heterogeneous workers have the ability to sort among ﬁrms whose compensation package best matches their preferences and ﬁrms have the incentive to tailor their compensation package to attract workers with desired characteristics (Hirth, Baughman, Chernew, & Shelton, 2006; Monheit &Vistnes, 1999). Rosen’s (1986) analysis of compensating differentials grounded research on the ﬁrm’s decision to offer beneﬁts to workers as a tradeoff with wages. Proﬁt maximizing ﬁrms must trade monetary and nonmonetary beneﬁts in compensating workers to keep costs competitive and utility maximizing workers will trade beneﬁts for wages. Individual-level negotiations of compensation might lead to wide variations in the composition of a ﬁrm’s compensation package to homogeneous workers as long as ﬁrms remain unconstrained, transaction costs are zero, worker preferences for the form of compensation vary, and workers sort among ﬁrms based on beneﬁts. However, because rules constrain a ﬁrm with respect to individual-level negotiation of compensation and create economies of scale when a ﬁrm offers a particular beneﬁt, the ﬁrm’s decision to offer a beneﬁt becomes inframarginal, with a low marginal cost of extending the offer to a large pool of workers once the decision to offer is made.

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

91

This research assesses the behavioral manifestations of inframarginal decision making in the ﬁrm’s offer of health insurance, one of the most critical components of a fringe beneﬁt package when workers sort among ﬁrms (Lehrer & Pereira, 2007). We argue that recruiting difﬁculties in positions other than the one in which compensation is set and the proportion of highskilled workers in the ﬁrm quantify manifestations of the ways in which rules and costs inﬂuence a ﬁrm’s decision to offer health insurance. We support their inﬂuence using the California Health and Employment Surveys (CHES), a database of 1,427 ﬁrms and 2,673 jobs in 27 Northern California and show how their omission in estimations of the wage-insurance tradeoff produces downward biases in the estimated wage-health insurance tradeoff.

FRAMEWORK A ﬁrm’s offer of health insurance is heavily regulated. Some legislation and rules are designed to incentivize a ﬁrm to offer insurance by lowering its cost. Under current law, employers may deduct the expenses they incur for employees’ health insurance and health care as a business expense (as they do for wages) but are exempt from paying the 6.2 percent payroll tax for Social Security (for workers below its maximum wage) and the 1.45 percent payroll tax for Medicare. These exemptions present the ﬁrm with a tax savings over wage compensation if it offers health insurance in lieu of wages. Still other rules are designed to ensure equity in coverage, with these rules generally raising the cost to the ﬁrm of offering health insurance. Many state mandates and some provisions of the newly legislated Affordable Care Act (ACA) require coverage of speciﬁc treatments or conditions. Section 105(h) of the Internal Revenue code prohibits discrimination in the offering of health insurance for ﬁrms with self-insured health plans1 (EBRI, 2009a) and cafeteria plans2 that include health insurance (nondiscrimination rule). The general idea behind the nondiscrimination rule is that beneﬁts to higher-paid employees must be equivalent to those offered to lower-paid employees. For a plan to be nondiscriminatory it must pass one of the three coverage tests: 70 percent of all employees beneﬁt under the plan; the plan beneﬁts 80 percent of eligible employees and 70 percent of all employees are eligible; and the plan beneﬁts a nondiscriminatory classiﬁcation of employees (e.g., the same type and level of beneﬁts are available to all). Firms offering multiple medical plans are unlikely to pass either the ﬁrst or second test because employees are likely to be dispersed among the various options, hence most ﬁrms satisfy the nondiscrimination rule by offering the same

92

NAN L. MAXWELL

plan to all eligible workers. This nondiscrimination rule affects a relatively large percentage of workers, albeit those employed in large ﬁrms. In 2008, 89 percent of workers employed in ﬁrms with 5,000 or more employees were in self-insured plans, leaving 55 percent of workers with health insurance coverage through a self-insured plan (EBRI, 2009b).3 When ﬁrms offer health insurance through a third-party, they are not affected by the nondiscrimination rule, however, many of the same restrictions arise from the contractional conditions the insurer requires to provide insurance. Because third-party insurers actively work to reduce the probability of adverse selection in the pool of individuals they cover (i.e., attracting a disproportionate share of people in poor health into their pool), they frequently impose many of the same minimum coverage restrictions contained in the nondiscrimination rule. Their interest lies in creating a group of workers that is not formed for health-related purposes because such a group is more likely to approximate a random sampling of the population, be actuarially sound, carry a lower probability of adverse selection, and have predictable expected costs of health care that are relatively stable over time. Insurers frequently structure their contracts to reward large, nonselect groups with explicit conditions on eligibility for coverage and lower premium prices for larger groups.4 The consequence of these rules is that the startup cost of offering insurance for a ﬁrm is high because it must meet conditions set by the nondiscrimination rule and third-party insurers. Economies of scale arise because larger groups face lower per-worker administrative costs for insurance (i.e., marginal cost is declining). Thirdparty insurers beneﬁt from contracting with a single, large ﬁrm in communicating with only a single individual or department, claims processing, beneﬁts administration, and medical underwriting. Insurance brokers commissions therefore consume 9 percent of the premium dollar for small ﬁrms but less than 0.1 percent for large ﬁrms (Yegian, 1999). Economies of scale also arise from declining administrative costs within the ﬁrm from less expensive and more competent governance structures in larger ﬁrms (Williamson, 1975) and the gains from specialization in purchasing and administering health beneﬁts. These externally and internally developed economies of scale in offering health insurance decrease its cost as ﬁrm size increases. As a result, administrative costs consume nearly 40 percent of every premium dollar for ﬁrms with 1–4 employees and about 5.5 percent for employers with 10,000 or more employees (as cited in Yegian, 1999). Administrative cost structures are also impacted with the number of plans offered (Moran, Chernew, & Hirth, 2001). As a ﬁrm increases the number of plans offered it must communicate the details of different plans to workers,

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

93

negotiate with different insurers, and face higher load charges and potentially higher premiums as the pool of workers in any one plan is reduced. In short, forces within the ﬁrm align with incentives provided by thirdparty insurers to create relatively high ﬁxed costs to offering insurance and relatively low marginal costs of expanding the pool of workers covered. The large ﬁxed costs of initially offering insurance suggest that proﬁt maximizing ﬁrms will offer health insurance only if the perceived beneﬁts are high. One potentially high set of beneﬁts is attracting workers with needed skills. Because workers sort across ﬁrms based on their preferences for health insurance (Goldstein & Pauly, 1976), ﬁrms can structure the health insurance offer to meet the preferences of workers with needed skills. This strategy works if preferences for wage and nonwage compensation vary by skills, a premise supported by research showing a correlation between wages, which are associated with skills, and preferences for nonwage compensation. Low-wage workers often prefer wages to nontaxed health beneﬁts (Currie & Yelowitz, 2000) as they often have a Medicaid safety net and low marginal tax rates. In contrast, high-wage workers often prefer nontaxed health beneﬁts to wages, at the margin, as they face high marginal tax rates.5 If a ﬁrm has a workforce with mixed preferences, research suggests it bases its offer on the preferences and prices faced by its median workers (Goldstein & Pauly, 1976). Highly compensated workers have a heavy inﬂuence on the offer (Gruber & Lettau, 2004), perhaps because ﬁrms want to attract skills and realize their loss by not offering insurance to highwage workers. Using health beneﬁts as a means to attract workers with needed skills might be especially appealing when ﬁrms face bottlenecks in hiring. Just as a ﬁrm might increase wages to attract workers with needed skills, it might also decide to offer health insurance to increase compensation, especially if the hiring difﬁculty lies in recruiting high-skilled workers, who have a strong preference for nonwage compensation. Although both worker preferences and recruiting difﬁculties in areas of labor bottlenecks might entice a ﬁrm to bear the relatively high ﬁxed cost and offer health insurance, its low marginal cost of extending the offer to a large pool of workers suggests the ﬁrm might extend the offer to workers throughout the ﬁrm once it decides to offer it. This inframarginal decision making means that workers who might not value an offer of health insurance highly will have an increased probability of receiving an offer simply because a high proportion of workers that value it highly are employed in the ﬁrm. Supporting this argument is research showing a pronounced incongruence between the offer of health insurance and worker preferences for high-wage workers in low-pay ﬁrms and low-pay workers in

94

NAN L. MAXWELL

high-pay ﬁrms (Carrington, McCue, & Pierce, 2002). Workers that receive an offer of health insurance but do not value it highly might not be willing to trade it for wages, however, and ﬁrms, who face a relatively low cost of offering it, might not reduce wages for fear of not being able to attract workers. Under such circumstances, the negative tradeoff between wages and health insurance would not be shown in empirical estimations. This research examines how the behavioral manifestations of the high ﬁxed and low marginal costs that a ﬁrm faces in offering health insurance – recruiting difﬁculties and workers preferences – might alter a ﬁrm’s probability of offering health insurance and produce a downward bias in the estimated work-insurance tradeoff. Speciﬁcally, we argue that the inframarginal decision to offer health insurance leads to two empirically testable propositions. First, recruiting difﬁculties anywhere in the ﬁrm, particularly among high-skilled workers, and an increasing percentage of high-skilled workers will increase the probability that a ﬁrm offers health insurance. The focus on high-skilled workers arises from their greater desire to trade wages for insurance and more critical role in production than workers with lesser skills (e.g., the lack of availability of production substitutes decreases as skill levels increase). As a result, rules and costs will incentivize ﬁrms to offer insurance in positions other than the one with recruiting difﬁculty and ﬁrms with a high proportion of high-skilled workers will be more likely to bear the cost of offering insurance with a low marginal cost extending it to additional workers who may not value it highly. Second, the estimated tradeoff between wages and health insurance will become stronger (or less positive) with controls for the inframarginal factors that increase the probability of a ﬁrm offering health insurance. Although compensating differentials predicts that ﬁrms will trade wages and health insurance in compensating workers, both rules and the low marginal cost of offering insurance may negate this tradeoff at the margin for it costs the ﬁrm little to extend the offer once the decision to offer it has been made. The tradeoff for the ﬁrm, at the margin, may therefore be close to zero for many workers, which would be veriﬁed as controls for the initial offer decision are included in empirical estimations.

RESEARCH DESIGN Estimations of whether a ﬁrm’s offer of health insurance, particularly those estimating a tradeoff with wages, are often confounded by unobserved

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

95

(to the researcher) individual, ﬁrm, and job heterogeneity that inﬂuence both the offer of health insurance and wages. One approach in controlling such heterogeneity is through the design of the research. Researchers have controlled for unobserved heterogeneity with experimental designs,6 differencing with multiple individual-level observations over time (e.g., Smith & Ehrenberg, 1983), and modeling individual variation in demand for health insurance through intrafamily trades (Olson, 2002). Of note, studies have been grounded in individual not ﬁrm-level estimations. This research takes a slightly different approach and uses ﬁrm and joblevel data collected from establishments7 to control for unobserved heterogeneity. Firm-level data capture the offer of health insurance at the appropriate decision-making level instead of relying on individual-level data to approximate the ﬁrm’s decision, as does the vast majority of empirical studies of the wage-health insurance tradeoff.8 We minimize ﬁrm-level heterogeneity using separate estimations on different samples of jobs drawn from the same sample of ﬁrms, which stands in contrast to using three different samples of ﬁrms for three skills groups. This more typical approach produces a fair amount of ﬁrm-speciﬁc unobserved heterogeneity. Heterogeneity in skill requirements is minimized by examining starting wages in jobs at each of the three narrowly deﬁned skill position levels in the ﬁrm:9 ‘‘low-skilled’’ (requiring a new worker to have no more than a high school education and no more than a year of work experience), ‘‘midskilled’’ (requiring some college or one to three years of work experience), and ‘‘high-skilled’’ (requiring at least a college degree and/or extensive work experience). Unobserved heterogeneity in the wage-insurance tradeoff is further reduced by using starting wages within each relatively narrow category of positions. Starting wages contain less heterogeneity than current wages because they occur at the start of the employment relationship when employers are largely ignorant of the productivity of new hires. This ignorance makes starting wages based less on observed productivity, training, and learning-by-doing than current wages and makes the variance in worker productivity in the employer’s eyes relatively small. As a result, much of the unobserved worker heterogeneity that determines wages and plagues econometric estimation of wage-insurance regressions is mitigated when using starting wages. Despite these design advantages, skill-based unobserved heterogeneity in estimations might exist in our three samples of jobs, particularly as increased education and work experience requirements help differentiate worker productivity (i.e., in mid and high-skilled jobs).

96

NAN L. MAXWELL

DATA AND EMPIRICAL ANALYSIS We used the California Health and Employment Surveys (CHES) to test our propositions. CHES is a cross section of 1,427 ﬁrms in 27 Northern California counties. Randomly selected ﬁrms were called from July 2005 through September 2006 with a 67-percent response (Maxwell, 2007). Firms with at least 51 employees were oversampled, with weights developed that could be used to ensure that analysis reﬂects the distribution of ﬁrms in the United States with respect to size and industry. The targeted respondent for answering the survey was ‘‘the person with knowledge about beneﬁts and jobs.’’ Government agencies and ﬁrms with fewer than ﬁve workers were excluded from the sampling frame. Firm-level data in the CHES include whether it offers health insurance, the percentage of workers employed in jobs at each skill level, and ﬁrm characteristics including industry, location, nonproﬁt status, and number of employees.10 Job-level data include information on the typical job at each skill level, including the ﬁrm’s difﬁculty in recruiting workers with needed skills,11 average starting wage and whether the incumbent worker is represented by collective bargaining. These data allow us to extend a tradition model (e.g., Claxton, DiJulio, Finder, & Becker, 2007; Glied & Zivin, 2004; Leibowitz & Chernew, 1992) of whether a ﬁrm (f) offers health insurance (H) as a function its characteristics (C: size, presence of a union, for proﬁt, rural location, and industry) to include a vector of variables that capture difﬁculty in recruiting workers (R) with needed skills in the typical job at each skill level (s) and a measure of the percentage of workers in high-skilled jobs in (HS):12 H f ¼ a0 þ Rs ar þ a1 HSf þ Cf ac þ 1

(1)

where e is a disturbance term. Of note, C in this estimation includes size and union variables that might be considered measures of institutions or costs. We classify them as ﬁrm characteristics in our equations because they are included in traditional models estimating the ﬁrm’s offer and we wish to focus our study on extensions to this model (i.e., R and HS). Estimation of Eq. (1) without the extensions (i.e., ar ¼ a1 ¼ 0) provides a benchmark of the estimated coefﬁcient size on ﬁrm characteristics (ac) on the probability of offering health insurance in traditional models. These benchmark coefﬁcients can be compared to those obtained when our measures of recruiting difﬁculty and workforce skills are included to determine biases with their exclusion. Our ﬁrst proposition posits that recruiting difﬁculties and an increasing percentage of high-skilled workers in the ﬁrm will increase the probability

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

97

that a ﬁrm offers health insurance, which we initially support with positive and signiﬁcant (pr.05) coefﬁcients on the measures of recruiting difﬁculty (ar) and high-skilled workers (a1) in Eq. (1) estimations. A more robust test will use job-level data to estimate the same general model outlined in Eq. (1) for jobs at each skill level. In this analysis we examine whether recruiting difﬁculties reﬂect market-based forces (i.e., increased compensation to attract workers with needed skills) or rules and cost structures that incentivize a ﬁrm to offer insurance to workers in jobs not necessarily affected by recruiting difﬁculties. We estimate the offer of health insurance to workers at each skill position level as a function of recruiting difﬁculties in the job at that skill level in which compensation is negotiated (su), which reﬂects market forces, and recruiting difﬁculties in jobs at other skill position levels (Rs–Rsu or Rus), which reﬂects rules and economies of scale that increase the probability of offering health insurance to workers in jobs outside the one in which the ﬁrm has difﬁculty in recruiting. We estimate: H f ;s ¼ a20 þ a21 HSf þ R0s a2r þ a22 Rs0 þ Cf a2c þ 2

(2)

using the same general interpretations and step approach in estimations described earlier. The second proposition posits that the estimated tradeoff between wages and health insurance will become stronger as controls for recruiting difﬁculty and high-skilled workers are included in the estimation. To test this proposition, we include the average starting wage13 (Ws) into Eq. (2) such that: H f ;s þ b0 þ b1 HSf þ R0s0 br þ b2 Rs0 þ Cf bc þ bw W s þ 3

(3)

Should bw become more negative (less positive) as Rus and HS are include into the model, our second proposition would be supported. We use ordinary least squares to estimate all models, which allows us to quantify the percentage increase in the probability of the ﬁrm making an offer of health insurance that is associated with a change in the independent variable.14 We use the unobserved heterogeneity in skill requirements that is increasingly present in higher-skilled jobs to perform a sensitivity analysis of our results. Because low-skilled jobs are relatively homogeneous with respect to skill requirements (i.e., require no more than a high school education and no more than one year of work experience), we expect skill differences that increase both wages and health insurance to be relatively minimal. Remaining heterogeneity would reside in ﬁrms and job attributes

98

NAN L. MAXWELL

(e.g., risk and capital usage) that determine productivity or compensation. However, as skill requirements increase (i.e., education or work experience), we expect heterogeneity to increase with variability in the nature of the skills acquired (i.e., as the sample moves from low- to high-skilled jobs), which will reduce estimated parameters in our models (a’s and b’s). We therefore expect propositions to have the strongest support when estimated on the sample of low-skilled jobs. The CHES data are uniquely suited for our study not only because they contain measures that capture recruiting difﬁculties and percent of highskilled jobs but also because they contain a sample of relatively typical ﬁrms and jobs. Of primary importance, ﬁrms in the CHES database show characteristics that likely subject them to the rules and economies of scale discussed earlier. Just over 5 percent are self-insured and just over 21 percent offer a cafeteria plan, with about 82 percent including health insurance in it (Table 1). These ﬁrms were subject to the nondiscrimination rule when data were collected. About 65 percent of ﬁrms have fewer than 20 workers and about 21 percent have 20–50 workers, which generally exposes them to experience rating and relatively high administrative costs in offering health insurance.

Table 1.

Rules, Costs, and CHES Firms.

Self-insured Firm offers cafeteria plan Health beneﬁt part of cafeteria plan Size 5–19 workers 20–50 workers 51–299 workers 300–999 workers 1000 þ workers Workers are all offered same beneﬁts Offer 1 plan only 5–19 workers 20–50 workers 51–299 workers 300–999 workers 1,000 þ workers

5.2% 21.1% 81.6% 65.2% 20.8% 8.6% 2.7% 2.7% 78.1% 50.1% 72.5% 17.2% 7.5% 1.8% 1.0%

Data Source: CHES (Maxwell, 2007). Observations were weighted so the distribution of ﬁrms reﬂects the proportion in the United States with respect to size and industry. Notes: Numbers represent the percentage of ﬁrm in the category.

99

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

The characteristics of jobs in the CHES suggest they might typify jobs at each of the skill position levels. Health insurance, hourly wage increase, recruiting difﬁculty, and the percentage of high-skilled workers all increase

Table 2.

Health insurance Tradeoff Wage at entry Market forces Recruiting difﬁculty in skill level Rules and costs Recruiting difﬁculty in high-skilled jobs Recruiting difﬁculty in mid-skilled jobs Recruiting difﬁculty in low-skilled jobs Percent high skilled Firm characteristics Large Union Rural For-proﬁt Manufacturing Service sector N

Descriptive Statistics. Low-Skilled Jobs

Mid-Skilled Jobs

High-Skilled Jobs

0.764 (0.403)

0.820 (0.369)

0.835 (0.359)

11.57 (4.99)

19.14 (11.00)

29.98 (15.65)

0.326 (0.445)

0.498 (0.480)

0.592 (0.476)

0.501 (0.475) 0.416 (0.469) –

0.532 (0.479) –

–

26.613 (23.418)

0.221 (0.398) 32.880 (25.430)

0.405 (0.476) 0.212 (0.396) 45.050 (29.203)

0.180 (0.365) 0.049 (0.203) 0.164 (0.352) 0.886 (0.301) 0.071 (0.244) 0.315 (0.441) 1,100

0.168 (0.359) 0.038 (0.184) 0.118 (0.309) 0.871 (0.321) 0.074 (0.250) 0.286 (0.433) 1,220

0.161 (0.356) 0.032 (0.171) 0.123 (0.318) 0.869 (0.327) 0.074 (0.254) 0.275 (0.432) 1,253

Data Source: CHES (Maxwell, 2007). Observations were weighted so the distribution of ﬁrms reﬂects the proportion in the United States with respect to size and industry. Notes: Numbers represent means and (standard deviations). Sample sizes vary with item-speciﬁc nonresponse. Most importantly samples fall to 891 for low-skilled jobs, 921 for mid-skilled jobs, and 861 for high-skilled jobs for wages.

100

NAN L. MAXWELL

with the skill requirements of the job (i.e., increase as the sample of jobs moves from low-skilled to high-skilled). About 76 percent of ﬁrms housing the sample of low-skilled jobs offer health insurance compared to about 83 percent of ﬁrms housing the sample of high-skilled jobs. Wages increase from an average of $11.57 in low-skilled jobs to $29.98 in high-skilled jobs and difﬁculty in attracting workers with needed skills increases from about 33 percent in low-skilled jobs to 59 percent in high-skilled jobs. The percentage of high-skilled workers in the ﬁrm increases from about 27 percent in the sample of low-skilled jobs to about 45 percent in the sample of high-skilled jobs (Table 2). CHES data provide evidence that heterogeneity increases as skill position levels increase. Both the coefﬁcient of variation for wages and the coefﬁcient of determination (R2) of the traditionally speciﬁed Eq. (2) estimation (ar ¼ a1 ¼ a2 ¼ 0) of health insurance at each skill position level can be used to assess unobserved heterogeneity in each of our samples of jobs. The CHES data show the lowest coefﬁcient of variation for wages and highest R2 in low-skilled jobs, as would be the case if unobserved heterogeneity in skills requirements increased with the skill requirements of the job (Fig. 1). Because heterogeneity may increase with different wage-to-productivity 0.700 0.600 0.500 0.400 Low Skilled 0.300 0.200

Mid Skilled High Skilled

0.100 0.000 Coefficient R square: offer varariation: wages

Fig. 1. Heterogeneity in Wages: Within each Skill Position Level. Data Source: CHES (Maxwell, 2007). Observations were Weighted so the Distribution of Firms Reﬂects the Proportion in the United States with respect to Size and Industry. Notes: Numbers Represent the Coefﬁcient of Variation for Wages (see Table 2 for Means and Standard Deviations) or the R2 from Estimating the Traditional Model of a Firm’s Decision to Offer Health Insurance (i.e., Without Recruiting or HighSkill Measures). Estimations are Not Shown, But Available from the Author.

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

101

Fig. 2. Heterogeneity in Wages: Coefﬁcient of Variation in Firms that Offer and Do Not Offer Health Insurance. Data Source: CHES (Maxwell, 2007). Observations were Weighted so the Distribution of Firms Reﬂects the Proportion in the United States with respect to Size and Industry. Notes: Numbers Represent the Coefﬁcient of Variation for Wages in Firms that Offer Health Insurance and Those That Do Not.

relationships between ﬁrms that offer and do not offer health insurance, we examine the coefﬁcient of variation for wages separately for each type of ﬁrm (Fig. 2). Firms that offer health insurance have a lower coefﬁcient of variation at each skill level than ﬁrms that do not offer it, as might be expected since compensation is more easily negotiated at the individual level without health insurance and a lower level of unionization. In ﬁrms that offer health insurance, the coefﬁcient of variation rises with skill level, as expected. In ﬁrms not offering health insurance, the coefﬁcient is highest in mid-skilled jobs and slightly higher in high-skilled than low-skilled jobs, which suggests a skill compression among workers in high-skilled jobs in ﬁrms that do not offer health insurance.

RESULTS As evidence that ﬁrms behave in accordance with the incentives provided by rules and economies of scale, CHES data show that about half the ﬁrms offer workers one health insurance plan only (Table 1). The percentage varies dramatically with ﬁrm size, as our framework suggests, with about 72 percent

102

NAN L. MAXWELL

of ﬁrms with fewer than 20 workers and 1 percent of ﬁrms with 1,000 or more workers offering only one plan. CHES data also suggest that 78 percent of ﬁrms offer the same beneﬁts to all workers. Most ﬁrms (about 51 percent) cited hours or tenure work requirements as the way in which beneﬁts differed (data not shown). No ﬁrm that differed in the beneﬁts offered explicitly mentioned differences in health plans as the demarcation of different beneﬁts, although about 15 percent mentioned that they varied beneﬁts with union representation and about 19 percent mentioned differences along the lines of position level. Only about 10 percent of these ﬁrms explicitly mentioned that health beneﬁts differed across job classiﬁcations, suggesting that only about 7.8 percent of ﬁrms differentiate their health insurance offer in ways that are inconsistent with the discussion in our framework. Our multivariate estimations using CHES data also provide support for both of our propositions about rules and costs. Firm-level estimations of Eq. (2) show that recruiting difﬁculty in any skill position level and the percentage of high-skilled workers in the ﬁrm increases the probability the ﬁrm will offer health insurance (Table 3), as predicted by our ﬁrst proposition. The coefﬁcients on recruiting difﬁculty are sizable and increase with the skills needed in the job. If a ﬁrm has difﬁculty recruiting appropriately skilled workers in high-skilled positions, it has about a 13 percentage point increased probability of offering health insurance, while a ﬁrm with recruiting difﬁculties in low-skilled positions has only a 5.7 percentage point increase (Table 3, column 2). Although the coefﬁcient on the percentage of high-skilled workers seems relatively small (0.002), a 20 percentage point increase in the percent of high-skilled jobs in the ﬁrm – roughly the difference between our samples of low- and high-skilled jobs – produces a 4 percentage point increase in the probability of a ﬁrm offering health insurance. Of note, this inﬂuence is signiﬁcant only for workers in low- and mid-skilled positions.15 Our more robust test of the ﬁrst proposition uses job-level data to examine the two distinct components of the recruiting difﬁculty measures: Rsu, which measures the market-force inﬂuence of increasing compensation with recruiting difﬁculty in the job, and Rus, which measures the inﬂuence of rules and economies of scale on health insurance with recruiting difﬁculty in jobs in which compensation is not being set. Results of these estimations suggest that market forces (along the diagonal in Table 3) increase the probability of offering health insurance by about ﬁve to six percentage points, irrespective of the skill position level. Recruiting difﬁculty in positions other than the one in which compensation is negotiated (rules and costs) has a stronger inﬂuence if the difﬁculty lies in recruiting workers in

103

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

Table 3.

Rules and Costs, Markets, and a Firm’s Offer of Health Insurance. Offer Health Insurance Traditional Model All Low-Skilled Mid-Skilled High-Skilled for All Firms Firms Jobs Jobs Jobs

Markets, rules, and costs Recruiting difﬁculty, high-skilled jobs Recruiting difﬁculty, mid-skilled jobs Recruiting difﬁculty, low-skilled jobs % High skilled Firm characteristics Large Union For-proﬁt Rural Service sector Manufacturing Intercept R2 N

–

0.129

0.114

0.128

0.052

–

0.093

0.101

0.052

0.133

–

0.057

0.062

0.025

0.027

–

0.002

0.002

0.002

–0.000

0.186 0.115 –0.147 –0.148 –0.058 0.150 0.908 0.080

0.185 0.137 –0.128 –0.143 –0.042 0.143 0.715 0.152

0.204 0.124 –0.131 –0.163 0.014 0.136 0.676 0.184

0.162 0.120 –0.117 –0.152 –0.028 0.113 0.762 0.129

0.140 0.151 –0.109 –0.246 –0.047 0.084 0.865 0.141

1,399

1,358

1,052

1,165

1,194

Data Source: CHES (Maxwell, 2007). Observations were weighted so the distribution of ﬁrms reﬂects the proportion in the United States with respect to size and industry. Notes: Numbers are unstandardized coefﬁcients from OLS estimations of text Eqs. (1) and (2) with health insurance as the dependent variable. Bold indicates signiﬁcant (pr0.05) coefﬁcients. – indicates that a variable was not included in model estimation.

mid- or high-skilled positions. Firms having difﬁculty recruiting appropriately skilled workers in mid- or high-skilled jobs have a 10 and 13 percentage point increase in the probability of offering health insurance.16 If, however, a ﬁrm has difﬁculty recruiting in low-skilled jobs, the probability of workers in mid- and high-skilled jobs receiving an offer of health insurance is not signiﬁcantly increased. Estimations also support our second proposition that the estimated tradeoff between wages and health insurance becomes stronger with controls for recruiting difﬁculties and workforce composition. We tested this proposition by estimating two versions of Eq. (3): the traditional tradeoff model in which the measures of recruiting difﬁculties and percent of high-skilled

104

NAN L. MAXWELL

workers are not included, and a model that includes these measures. The traditional estimation on the sample of low-skilled jobs (Table 4) shows a positive and signiﬁcant (pr0.05) wage coefﬁcient consistent with much of the empirical research on the wage-insurance tradeoff. As we add the variable that captures the recruiting difﬁculty in the job (Market Forces column), the wage coefﬁcient becomes insigniﬁcant but remains positive. As we add variables that capture the recruiting difﬁculty in other skill position levels, the wage coefﬁcient becomes negative but insigniﬁcant. These results support the proposition that a failure to control for a ﬁrm’s behavioral manifestations of accompanying rules and economies of scale might contribute to the failure of empirical research to support a tradeoff between wages and health insurance. These results break down somewhat as Table 4.

Institutions, Markets, and a Firm’s Offer of Health Insurance in Low-Skilled Jobs.

Tradeoff Wage Market forces Recruiting difﬁculty in skill level Rules and costs Recruiting difﬁculty, high-skilled jobs Recruiting difﬁculty, mid-skilled jobs Recruiting difﬁculty, low-skilled jobs % High skilled Firm characteristics Large Union For-proﬁt Rural Service sector Manufacturing Intercept R2 N

Traditional Tradeoff Model

Market Forces

Rules and Costs

0.006

–

–0.001

–

0.119

0.065

–

–

0.095

– – –

– – –

0.133 – 0.002

0.216 0.043 –0.150 –0.156 –0.024 0.123 0.813 0.103

0.223 0.058 –0.165 –0.178 –0.041 0.118 0.818 0.119

0.210 0.121 –0.140 –0.183 –0.007 0.100 0.707 0.181

880

877

855

Data Source: CHES (Maxwell, 2007). Observations were weighted so the distribution of ﬁrms reﬂects the proportion in the United States with respect to size and industry. Notes: Numbers are unstandardized coefﬁcients from OLS estimations of text Eq. (3) with health insurance as the dependent variable. Bold indicates signiﬁcant (pr0.05) coefﬁcients. – indicates that a variable was not included in model estimation.

105

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

we increase the heterogeneity in skill requirements of the job and estimate equations on the samples of mid- and high-skilled jobs (Table 5). For these samples, the wage coefﬁcient is insigniﬁcant in all estimations, although it moves from positive to negative in the sample of mid-skilled jobs when measures of recruiting difﬁculty in high- and low-skilled jobs and the percentage of high-skilled workers are included. It remains positive and insigniﬁcant in the sample of high-skilled jobs in all model speciﬁcations.

Table 5.

Adding Heterogeneity: Mid- and High-Skilled Jobs. Mid-Skilled Jobs Traditional Tradeoff Model

Tradeoff Wage Market forces Recruiting difﬁculty in skill level Rules and costs Recruiting difﬁculty, high-skilled jobs Recruiting difﬁculty, mid-skilled jobs Recruiting difﬁculty, low-skilled jobs % High skilled Firm characteristics Large Union For-proﬁt Rural Service Manufacturing Intercept R2 N

High-Skilled Jobs

Market Rules and Forces Costs

Traditional Tradeoff Model

Market Rules and Forces Costs

0.001

0.000

–0.000

0.001

0.001

0.001

–

0.142

0.110

–

0.085

0.037

–

–

0.096

–

–

–

–

–

–

–

–

0.111

–

–

–0.012

–

–

0.028

–

–

0.001

–

–

–0.001

0.157 0.110 –0.116 –0.133 0.004 0.093

0.177 0.123 –0.132 –0.152 0.013 0.086

0.171 0.136 –0.110 –0.160 0.016 0.082

0.120 0.121 –0.069 –0.150 0.012 0.089

0.127 0.147 –0.071 –0.157 0.011 0.082

0.127 0.159 –0.086 –0.188 0.005 0.071

0.889 0.066

0.834 0.101

0.769 0.119

0.865 0.065

0.810 0.079

0.837 0.113

915

906

889

854

839

834

Data Source: CHES (Maxwell, 2007). Observations were weighted so the distribution of ﬁrms reﬂects the proportion in the United States with respect to size and industry. Notes: Numbers are unstandardized coefﬁcients from OLS estimations of text Eq. (3) with health insurance as the dependent variable. Bold indicates signiﬁcant (pr0.05) coefﬁcients. – indicates that a variable was not included in model estimation.

106

NAN L. MAXWELL

Our sensitivity analysis suggests that the coefﬁcient estimates on ﬁrm characteristics (ac) are fairly robust with respect to the level of unobserved job heterogeneity in skills while estimates of recruiting difﬁculties (ar) are sensitive to it. We tested for this robustness by estimation Eqs. (1) and (2) in stages: a traditional tradeoff model of the offer of health insurance in which the measures of recruiting difﬁculties and percent of high-skilled workers are not included and a model that includes both measures (Table 3). When the recruiting difﬁculty and percentage of high-skilled workers were included in the estimation, few differences arise in the size or signiﬁcance of the coefﬁcients on the ﬁrm characteristic variables (C), suggesting that models estimated without their inclusion might not produce biased estimates of the relationship between the ﬁrm’s characteristics and its decision to offer health insurance. Estimates of the relationship between our measures of behavioral manifestations of rules and economies of scale – recruiting difﬁculty in particular – appear to be inﬂuenced by the level of unobserved job-related heterogeneity, however (Table 3). The estimated coefﬁcients from the sample of low-skilled jobs – our most homogeneous sample with respect to skill heterogeneity – are about the same size as for the sample of ﬁrms. However, results change when estimations are performed on the sample of mid- and high-skilled jobs. Recruiting difﬁculties in low-skilled jobs do not signiﬁcantly increase the probability that a ﬁrm will offer health insurance in either sample and the inﬂuence of high-skilled workers is not signiﬁcantly related to the probability of it offering health insurance in the sample of high-skilled jobs.

SUMMARY AND DISCUSSION We argued that the rules and economies of scale that ﬁrms face when offering fringe beneﬁts create high ﬁxed but low marginal costs in offering it and, as a result, move the wage-beneﬁt tradeoff for the ﬁrm away from the individual level. We ﬁnd empirical support for the argument using the case of health insurance. The nondiscrimination rule for ﬁrms that are self-insured or offer a cafeteria plan that includes health insurance, the conditions third-party insurers put on the coverage, and administration costs all create economies of scale in a ﬁrm’s offering health insurance and, as a consequence, create incentives for the ﬁrm to make inframarginal decisions as to whether it offers workers fringe beneﬁts. One potentially large advantage that ﬁrms gain from bearing the high ﬁxed cost of offering insurance is attracting workers with needed skills. Because workers sort across ﬁrms based on their preferences for health insurance, ﬁrms can use their offer of health insurance to attract

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

107

workers with desired skills. Once the ﬁrm’s decision to offer health insurance is made, the low marginal cost and institutional rules provide incentives to extend the offer to a large pool of workers. We argue that ﬁrms are likely to beneﬁt from (and hence bear the high ﬁxed cost of) offering health insurance as the percentage of high-skilled jobs increases because high-skilled workers have a stronger preference for nonwage compensation than low-skilled workers. We also argue that the low marginal cost of extending the offer to additional workers suggests that workers in low- and mid-skilled positions will have an increased probability of receiving an offer of health insurance with an increasing presence of highskilled workers in the ﬁrm. We also argue that difﬁculty in recruiting workers provides another incentive for ﬁrms to bear the high ﬁxed cost of offering health insurance, especially if the difﬁculty is in recruiting workers for highskilled positions, with the relatively low marginal cost and institutional rules increasing the probability of workers offered health insurance in positions outside the one in which the compensation is negotiated. Our analyses of ﬁrm and job-level data in the CHES support these arguments. We ﬁnd that a 20 percentage point increase in the percentage of high-skilled workers in the ﬁrm – roughly the percentage difference between our samples of low- and high-skilled jobs – produces a four percentage point increase in the probability of a ﬁrm offering health insurance. We also show a 13 percentage point increase in the probability of a ﬁrm offering health insurance if it has difﬁculty recruiting in high-skilled positions and a 5.7 percentage point increase with recruiting difﬁculties in low-skilled positions. When we compare the recruiting difﬁculty in the job for which compensation is negotiated (market forces) with the recruiting difﬁculty in positions other than the one in which compensation is negotiated (rules and economies of scale), we ﬁnd a larger increase in the probability of an offer when difﬁculties arise outside the area of compensation negotiation, as long as the difﬁculty is in mid- or high-skilled positions. Recruiting difﬁculty in positions in which compensation is being negotiated increases the probability of offering health insurance by about ﬁve to six percentage points, irrespective of the skill position level with the percentage increase about 10–13 points if the difﬁculty is in mid- or high-skilled positions other than the one in which compensation. One critical insight from this study is that it draws attention to the consequences that rules and economies of scale play in inﬂuencing a ﬁrm to offer health insurance. The relatively high ﬁxed cost that a ﬁrm bears in offering insurance and decreasing marginal costs that it faces may create inframarginal decision making in offering fringe beneﬁts and move the tradeoff with wages away from the individual level. Once the decision to

108

NAN L. MAXWELL

offer beneﬁts is made, the low marginal cost of extending the offer to additional workers may negate a tradeoff with wages for many workers, particularly those that do not value the beneﬁt, which presents problems in empirical establishing a wage-health insurance tradeoff (for example). By focusing on ﬁrm behavior in light of these considerations, our research shows how including factors that affect the probability of a ﬁrm’s offering health insurance, but not necessarily its tradeoff with wages at the individual level, helps to reduce the positive empirical relationship that has plagued microlevel analysis of the wage and health insurance tradeoff. Inframarginal decision making might be attractive from a policy perspective if it decreases the number of uninsured. If rules and economies of scale create incentives for ﬁrms to offer health insurance to low-wage workers who value the offer less than the equivalent offer of wages, the number of uninsured might be decreased. Such workers might not take up health insurance without the heavy subsidy of premium that an employer often pays and might not take a job offer if wages were lowered by the cost of insurance. Because rules and costs might provide ﬁrms with an incentive to extend an offer of insurance to such workers without reducing wages, a greater proportion of the population might be insured. The underlying impetus for the ACA extending the nondiscrimination rule to virtually all employer-based plans (allowing for some grandfathering) was likely grounded in such reasoning. Alternatively, extending the nondiscrimination rule may increase a ﬁrm’s ﬁxed cost of offering insurance, if the third-party insurers requirements for structuring the offer are less stringent than the nondiscrimination rule. Under such circumstances ﬁrms that do not offer insurance will bear higher start up costs in making an offer and some ﬁrms might face the possibility of dropping insurance with elevated costs. We leave to future research the question of which effect is strongest.

NOTES 1. A self-insured plan is one in which the ﬁrm acts as its own insurer and bears the risk of providing health coverage for insured events, even if the employer contracts with an insurance company to administer the plan. In contrast, fully insured plans are ones in which the ﬁrm pays a per-employee premium to a third party (insurance company), who then assumes the risk of providing health care for insured events. Before the ACA, Congress restricted the nondiscrimination rule to self-insured plans. 2. Beneﬁts received through a cafeteria plan are taxed because they are deemed to be in constructive receipt of the cash unless the plan meets various reporting and nondiscrimination requirements (Lyke, 2006).

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

109

3. The nondiscrimination rule provides several ways in which a ﬁrm can discriminate. It can establish separate plans for distinct classes of large groups (more than 50 employees) with a business rationale (e.g., hourly and salaried), although explicit grouping by compensation level is not allowed. It can also exclude certain workers from the rule (e.g., less than three years of tenure, under age 25, part time or seasonal, and under a collective bargaining agreement). Carrington et al. (2002) provide a discussion. 4. Insurers typically do not assess risk in ﬁrms with 500 or more workers but rate the smallest ﬁrms individually for risk, just as in the nongroup (individual) market, because a single sick employee with a high utilization level could inﬂict a loss on the insurer (Cutler, 1994). 5. All workers pay taxes on wages but not health insurance. However, workers with a 35 percent marginal rate face greater tax savings from tax-free beneﬁts than workers with a 5 percent marginal rate. Pauly (2001) provides a succinct overview. 6. See RAND’s Health Insurance Experiment (Newhouse, 1993) and the natural experiment with mandated maternity beneﬁts (Gruber, 1994). 7. The sampling frame for the data was establishments with multiple establishments in the same ﬁrm not included. 50.5 percent of the samples are single-establishment ﬁrms. We discuss the data as if the ﬁrm were the unit of analysis because health beneﬁts were virtually all set by the ﬁrm, not the establishment. Only 62 establishments in multi-establishment ﬁrms (4.3 percent) report setting their own beneﬁts. 8. Such analysis overstates the decision making of larger ﬁrms with their greater employment. 9. Respondents were asked ‘‘about the different types of positions you have (in the establishment). We are particularly interested in learning about positions requiring different levels of education and work experience. In answering these questions, we would like you to think about ALL the positions in this ﬁrm at this location and to classify them by the education and training level required of workers when they start the job.’’ 10. Firm size categories were structured such that ﬁrms could be delineated as to whether they were subjected to the rules governing the small group health insurance market in California (2–50 employees) (Roth, 2003). 11. Respondents used a four-point scale to assess their ability to attract workers with the needed skills. Jobs classiﬁed as ‘‘a little hard’’ or ‘‘very hard’’ to attract needed skills were categorized (¼1) as hard to attract skills in job. 12. Appendix Table A1 provides an empirical deﬁnition for each measure. We included county level unemployment in estimations as an additional control for county level differences. Estimated coefﬁcients were virtually identical to those presented in our tables and the variable gained signiﬁcance (at pr0.05) only in the ‘‘Traditional Model’’ and the ‘‘Mid-Skilled Jobs’’ model presented in Table 3. 13. The question asked ‘‘$____ per _____’’ (e.g. $9.85/h). We assumed full-time work when computing hourly wages for responses not reported as hourly in the raw data, (e.g., annual salary was divided by 2080, monthly salary was divided by 173, and weekly was divided by 40). About 92 percent of wages in low-skilled jobs were reported as hourly, compared to about 80 percent of mid-skilled jobs and 52 percent of high-skilled jobs. Conversely, about 42 percent of wages in high-skilled jobs were reported in annual salary, compared to about 19 percent in mid-skilled and 8 percent

110

NAN L. MAXWELL

in low-skilled jobs. Surveyors were instructed to remind respondents that the questions pertained to ‘‘when [workers] start the job’’ when asking the question about average wage in the job at skill position level. 14. We conﬁrm the OLS results with marginal effects estimated through probits (see appendix Tables A2 and A3). We present the OLS estimations because the probits estimated for the high-level jobs do not produce stable estimates of the coefﬁcient on the union variable. Only 72 of the high-skilled ﬁrms do not offer health insurance and none of them are unionized (over 90 percent are for proﬁt). Because unionization is too central to the literature on the ﬁrm offering health insurance to drop it from the model, we report the OLS estimations. 15. Firms with a higher percentage of high-skilled workers may have low-skilled workers who are of higher quality in unobservable ways, thus are receiving health insurance for that reason. We cannot identify these potential quality differences with the broad education categories used to deﬁne high- and low-skilled workers. 16. Difﬁculty in hiring might be correlated with the type of workers sought, with workers at the high end of the skill distribution harder to recruit at all levels of the skill distribution.

ACKNOWLEDGMENTS The research was undertaken while the author was a professor of Economics and director of the HIRE Center at California State University, East Bay and mostly ﬁnalized when she was a visiting research fellow at the Public Policy Institute of California. It beneﬁted greatly from comments of Steve Shmanske and Jed DeVaro and from presentations at the W.E. Upjohn Institute for Employment Research and the Western Economic Association. The research was partially funded by the W.E. Upjohn Institute for Employment Research and University of California Program on Access to Care. Although numerous individuals made this study possible, the author assumes full responsibility for all errors.

REFERENCES Carrington, W. J., McCue, K., & Pierce, B. (2002). Nondiscrimination rules and the distribution of fringe beneﬁts. Journal of Labor Economics, 20(Pt. 2), S5–S33. Claxton, G., DiJulio, B., Finder, B., & Becker, E. (2007). Employer health beneﬁts, 2007 (annual survey). Retrieved from http://www.kff.org/insurance/7672/upload/76723.pdf Cutler, D. M. (1994). Market failure in small group health insurance. Retrieved from http:// www.nber.org/papers/W4879.pdf Employee Beneﬁt Research Institute (EBRI). (2009a). Fundamentals of employee beneﬁt programs. Retrieved from http://www.ebri.org/publications/books/index.cfm?fa¼fundamentals Employee Beneﬁt Research Institute (EBRI). (EBRI). Health plan differences: Fully-insured vs. selfinsured. Fast Facts, Retrieved from http://www.ebri.org/pdf/FFE114.11Feb09.Final.pdf

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

111

Employee Beneﬁt Research Institute (EBRI). (2010). EBRI databook on employee beneﬁts. Retrieved from http://www.ebri.org/publications/books/index.cfm?fa¼databook Glied, S. & Zivin, J.G. (2004). Modeling employer decisions to offer health insurance. Retrieved from http://www.rwjf.org/ﬁles/research/no7researchabstract.pdf Goldstein, G. S., & Pauly, M. V. (1976). Group health insurance as a local public good. In: R. R. Rossett (Ed.), The role of health insurance in the health services sector (pp. 73–110). New York: NBER. Gruber, J. (1994). The incidence of mandated maternity beneﬁts. American Economic Review, 84(3), 622–641. Gruber, J., & Lettau, M. (2004). How elastic is the ﬁrm’s demand for health insurance. Journal of Public Economics, 88(7–8), 1273–1293. Hirth, R. A., Baughman, R. A., Chernew, M. E., & Shelton, E. C. (2006). Worker preferences, sorting and aggregate patterns of health insurance coverage. International Journal of Health Care Finance Economics, 6, 259–277. Lehrer, S. F., & Pereira, N. S. (2007). Worker sorting, compensating differentials and health insurance: Evidence from displaced workers. Journal of Health Economics, 26(5), 1034–1056. Leibowitz, A., & Chernew, M. (1992). The ﬁrms’ demand for health insurance. In: Health beneﬁts and the workforce (pp. 77–83). Washington DC: USGPO. Lyke, B. (2006). Tax beneﬁts for health insurance and expenses: Overview of current law and legislation. CRS Report for Congress. Retrieved from http://assets.opencrs.com/rpts/ RL33505_20060630.pdf Maxwell, N. L. (2007). The California health and employment surveys (CHES) technical report on methods. Retrieved from http://www.hire.csueastbay.edu/hire/discpap/abstracts/ R07-01-01.pdf Monheit, A. C., & Vistnes, J. P. (1999). Health insurance availability at the workplace: How important are worker preferences?. Journal of Human Resources, 34(4), 770–785. Moran,, J. R., Chernew,, M. E., & Hirth, R. A. (2001). Preference diversity and the breadth of employee health insurance options. Health Services Research, 36(5), 911–934. Newhouse, J. P. (1993). Free for all? Lessons from the RAND health insurance experience. Cambridge, MA: Harvard University Press. Olson, C. A. (2002). Do workers accept lower wages in exchange for health beneﬁts? Journal of Labor Economics, 20(Pt. 2), S91–S114. Pauly, M. V. (2001). Making sense of a complex system: Empirical studies of employmentbased health insurance. International Journal of Health Care Finance and Economics, 1(3/4), 333–339. Rosen, S. (1986). The theory of equalizing differences. In: O. Ashenfelter & R. Layard (Eds.), Handbook of labor economics (pp. 642–692). New York: Elsevier Science. Roth, D. L. (2003). Insurance markets: Rules governing California’s small group health insurance market. Retrieved from http://www.chcf.org/topics/healthinsurance/index. cfm?itemID¼20740 Smith, R. S., & Ehrenberg, R. G. (1983). Estimating wage-fringe trade-offs: Some data problems. In: J. E. Trippett (Ed.), The measurement of labor cost (pp. 347–367). Chicago, IL: University of Chicago Press. Williamson, O. E. (1975). Markets and hierarchies: Analysis and antitrust implications. New York: The Free Press. Yegian, J. M. (1999). Size matters: The health insurance market for small ﬁrms. Brookﬁeld, VT: Ashgate.

112

NAN L. MAXWELL

APPENDIX

Table A1.

Variable Deﬁnitions. Deﬁnition

Health insurance Health insurance Tradeoff Wage at entry (Ws) Market forces Recruiting difﬁculty in the skill position level (Rsu)

Rules and costs Recruiting difﬁculty not in a position in which compensation is negotiated (Rus)

Percent high skilled (HS) Firm characteristics (Cf) Large Union

For-proﬁt Rural Service

Manufacturing

A binary variable with 1 indicating the ﬁrm offers health beneﬁts and 0 indicating it does not Starting hourly rate of pay in a typical job in the skill position level (s) A binary variable with 1 indicating the ﬁrm has at least some difﬁculty recruiting workers with the required skills in the skill position level in which compensation is negotiated (su) and 0 indicating it does not A series of binary variables with 1 indicating the ﬁrm has at least some difﬁculty recruiting workers with the required skills in a skill position level in which compensation is NOT negotiated and 0 indicating it does not The percentage of high skilled (at least a college degree or 5 years of work experience) in the ﬁrm A binary variable with 1 indicating the ﬁrm has more than 50 employees and 0 indicating it does not A binary variable with 1 indicating that workers in the job or somewhere in the ﬁrm (in equation 1) are represented by a union and 0 indicating they are not A binary variable with 1 indicating the ﬁrm is for proﬁt and 0 indicating it is not A binary variable with 1 indicating the ﬁrm is located in a rural area and 0 indicating it is not A 0, 1 binary variable with 1 indicating a ﬁrm in the service sector (1987 SIC of 70–72, 74–79, 81, 83–86, 88–89) and 0 indicating it is not A 0, 1 binary variable with 1 indicating a ﬁrm in the manufacturing sector (1987 SIC of 20–39) and 0 indicating it is not

113

How Do Rules and Costs Affect a Firm’s Setting of Beneﬁts?

Table A2.

Marginal Effects: Probit Estimations Excluding Wages. Offer Health Insurance All Firms

Low-Skilled Jobs

Mid-Skilled Jobs

High-Skilled Jobs

0.126

0.100

0.122

0.048

0.090

0.094

0.053

0.131

Recruiting difﬁculty, high-skilled jobs Recruiting difﬁculty, mid-skilled jobs Recruiting difﬁculty, low-skilled jobs % High skilled

0.054

0.057

0.013

0.019

0.001

0.002

0.001

–0.000

N

1,358

1,052

1,165

1,194

0.104 – – – –

877

– – – – –

880

N

855

0.002

0.059 0.087 0.126

–0.002

Rules and Costs

915

– – – – –

0.001

Traditional Model

906

0.132 – – – –

0.000

Market Forces

889

–0.020 0.001

0.108 0.097

–0.001

Rules and Costs

854

– – – – –

0.001

Traditional Model

839

0.082 – – – –

0.001

Market Forces

834

0.038 – 0.111 0.015 –0.001

0.001

Rules and Costs

High-Skilled Jobs

Data Source: CHES (Maxwell, 2007). Observations were weighted so the distribution of ﬁrms reﬂects the proportion in the United States with respect to size and industry. Notes: Numbers represent the average marginal effects that are computed from a probit estimation of text Eq. (2) (Table A27) or Eq. (3) (Table A38) with health insurance as the dependent variable. Bold indicates signiﬁcant (pr0.05) coefﬁcients. designates probit estimations with instability in the estimated union coefﬁcient. – indicates that a variable was not included in model estimation. Coefﬁcients reported as signiﬁcant in OLS estimations, but not in probit estimations have the p value fall between 0.05 and 0.07 (i.e., .05opo.07).

0.003

0.005

Market Forces

Wage Recruiting difﬁculty in Job at skill level High-skilled jobs Mid-skilled jobs Low-skilled jobs % High skilled

Traditional Model

Mid-Skilled Jobs

Marginal Effects: Probit Estimations Including Wages.

Low-Skilled Jobs

Table A3.

114 NAN L. MAXWELL

MAJORITY OWNERSHIP AND CHIEF EXECUTIVE COMPENSATION Derek C. Jones and Niels Mygind ABSTRACT In this chapter, we provide the ﬁrst empirical study of the effects of differing types of majority ownership, including employee ownership, on executive compensation. By investigating the case of Estonia, we also extend the range of geographical coverage of studies of the determinants of executive compensations to the case of Estonia. Although previous research ﬁnds that the type of ownership affects CEO pay, our new panel data, and the exceptional conﬁgurations of ownership that prevailed in Estonia during early transition, enable us to construct unusual measures of majority ownership. Findings indicate that an economically signiﬁcant determinant of CEO pay is ownership both in state versus privatized ﬁrms and in different types of private ﬁrms. In ﬁrms with majority ownership by employees, pay is about 15% less than in state-owned ﬁrms, other things equal. CEO pay is also positively related to size and seldom related to performance although size elasticities are much smaller than those estimated in other studies, mainly for advanced western countries.

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 115–141 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012009

115

116

DEREK C. JONES AND NIELS MYGIND

Findings provide more general support than previously for the varying importance of principal–agency relationships across ﬁrm types and the views that privatization and employee ownership have imposed strong discipline on the level of CEO compensation. Keywords: Executive compensation; ownership structures; employee ownership; privatization; Estonia JEL classiﬁcations: J3; P00; I2

INTRODUCTION A recent survey (Frydman & Jenter, 2010) of the often highly charged subject of executive compensation notes that most early empirical work concentrated on the United States. The range of more recent empirical work has broadened substantially with studies including most developed market economies and some emerging economies. Perhaps surprisingly, apparently only two studies exist for the former Soviet economies, namely, Jones and Kato (1996) for Bulgaria and Eriksson (2005) for the Czech Republic. In this chapter, we extend geographical coverage by providing the ﬁrst empirical paper on executive compensation in Estonia. Our focus is on the consequences of different ownership arrangements for executive pay. Theorists predict that differences in ownership and organizational structure will be accompanied by variation in monitoring and incentive-alignment mechanisms with conﬂicting views on the role of managers’ ownership in affecting enterprise performance (e.g., compare Demsetz & Lehn, 1985; Jensen & Meckling, 1976). However, empirical work on propositions concerning ownership is limited in large part reﬂecting a narrow range of ownership structures in western economies. The available empirical work tends to focus on matters such as the effects of varying amounts of insider ownership on CEO pay and performance (e.g., Morck, Schleifer, & Vishny, 1988).1 Investigations of the effects on executive compensation of differences in majority ownership are quite scarce. Moreover, they tend to examine outcomes only for managers and make comparisons with benchmark conventional ﬁrms that have dispersed ownership. For example, Denis and Denis (1994) use data for US ﬁrms to investigate the links between majority ownership by managers and organizational efﬁciency.2

Majority Ownership and Chief Executive Compensation

117

The disintegration of the former Soviet economies led to adaptations of principal-agent theory to that case (e.g., Boycko, Schleifer, & Vishny, 1996). Among transition cases, the early Estonian experience was especially interesting because privatization produced a broad array of ownership types. In nearly all ﬁrms, a clear majority owner usually was apparent, and these differing majority ownership forms coexisted for several years. Thus, for a large stratiﬁed random sample, Jones and Mygind (1999) found that, in 1997, each of the following ﬁve groups were majority owners in at least 10% of sample ﬁrms – nonmanagerial employees, managers, foreigners, domestic outsiders, and the state.3 This ability to classify Estonian ﬁrms on the basis of majority ownership enables us to examine whether type of majority ownership is associated with CEO pay. Moreover, in recent years, some researchers (e.g., Bebchuk & Fried, 2004) have argued that executive compensation is most fruitfully examined through the lens of managerial power. It would seem that this theme is especially useful to examine during early transition, when insiders, notably managers, accumulated enormous power (e.g., Estrin, Hanousek, Koc, & Svejnar, 2009). The standard executive compensation literature identiﬁes key roles for size and enterprise performance in inﬂuencing CEO pay and also provides benchmark estimates of their economic importance (e.g., Murphy, 1999). It is instructive to see whether size and performance play similar roles in Estonia during early transition and thus help to gauge the extent to which the managerial labor market functioned during early transition.4 After brieﬂy reviewing key issues in the theoretical and empirical executive pay literature, we then relate some of these issues to the speciﬁc context facing former Soviet-type economies during early transition. After outlining basic aspects of the privatization process and transition in Estonia, we describe our unusual data – a representative panel of ﬁrms with corresponding data at the chief executive level. After outlining the empirical strategy, we report ﬁndings from our ﬁxed effects models.

CONCEPTUAL FRAMEWORK There is an extensive literature that predicts that differences in corporate ownership are expected to have links with ﬁrm performance and valuation including systematic relationships between ownership structure and pay (Frydman & Jenter, 2010; Murphy, 1999). In the main, this literature is concerned with publically held ﬁrms in which most ownership is held by outsiders (individuals and institutions that do not work in the ﬁrm) and in

118

DEREK C. JONES AND NIELS MYGIND

which ﬁrms are assumed to focus on proﬁtability and maximizing shareholder value. Much of this literature has concentrated on the effects of relatively limited differences in ownership by senior managers and how agency costs might be ameliorated by modest amounts of ownership by managers (e.g., Jensen & Meckling, 1976). By contrast, relatively little attention has been devoted to CEO pay in different types of ﬁrms and in which ﬁrms might pursue varying objectives.5 However, there are various reasons why CEO pay is expected to be different when ﬁrms are of quite different types. Consider ﬁrms in which the state is the majority owner (as remains the case in some ﬁrms in former communist countries, even after more than 20 years of transition). Some have argued (e.g., Aslund, 1999) that rent-seeking by managers in large state-owned ﬁrms can be expected to be more important than in privatized ﬁrms that are far less politicized and in which market forces will be expected to exert more discipline on CEO compensation.6 In addition, the special circumstances that confronted ﬁrms during early transition in former Soviet-type economies arguably introduce additional considerations that potentially exert greater force in ﬁrms that remained in the state sector. In an economy such as Estonia, in which hard budget constraints prevailed quite early in the reform process (OECD, 2000), the prediction of a strong link between executive pay and performance might be expected to be stronger than in countries with less ﬁnancial discipline. At the same time, the link is expected to be strong only if several other conditions were met including the existence of an active market for managers with a requirement that managers and other economic agents are well informed of comparative ﬁrm performance. When such assumptions are not met, the pay–performance link may be expected to be quite weak, as might be the link between size and performance for ﬁrms that started transition with bloated levels of employment. These links are expected to be especially weak in ﬁrms in which ﬁrm objectives moved only slowly away from objectives that ﬂourished under Soviet-type arrangements. For example, while for many private ﬁrms, proﬁts soon became the dominant goal during early transition; in many state ﬁrms, a higher value was likely to continue to be placed on other objectives such as employment stability. Furthermore, during early transition, the managerial labor market might display institutional legacies from Soviet times; managerial pay was apt to be a base wage and top managers would be paid a low multiple of the average wage. Theorists have pointed out how these arrangements would be expected to result in acute incentive and motivational problems for

Majority Ownership and Chief Executive Compensation

119

managers (e.g., Bonin, 1976). A substantial gap would emerge between behavior in a proprietary fashion (as called for under the ofﬁcial ideology) and the reality of risk aversion and the pursuit of a quiet life (Kornai, 1992). Partly because of asymmetries in information between managers and planners, a ‘‘ratchet effect’’ would emerge with extensive managerial slack (e.g., Ickes & Samuelson, 1987). In turn, several systemic inefﬁciencies were predicted (e.g., Putterman, 1993). Hence, to facilitate successful overall reform during early transition, many stressed the crucial importance of reforming incentive systems (e.g., Aghion, Blanchard, & Burgess, 1994). A need existed to structure executive pay to provide pecuniary incentives for managers to pursue proﬁtability and encourage more market-oriented managerial behavior. At the same time, the historical legacy and institutional inertia would act as a drag on the reform of executive pay systems. These forces might be expected to be stronger in ﬁrms that remained in the state sector and far weaker in ﬁrms in which new forces were more clearly evident, as in many privatized ﬁrms. As well as predicting differences in CEO pay between ﬁrms in the state versus the private sector, differences in the forces inﬂuencing CEO pay are also expected to vary according to the type of private ownership. For western ﬁrms, to the extent that this issue has been engaged, the literature has concentrated on agency costs and the effects of relatively limited differences in ownership by senior managers. The early literature (e.g., Jensen & Meckling, 1976) suggests that managerial ownership might be beneﬁcial for ﬁrms helping to better align the interests of managers with those of dispersed outside shareholders. But later work argued that considerable costs were thought to be associated with large managerial ownership stakes with powerful managers entrenching themselves, obstructing board monitoring activities, and introducing policies (including excessive compensation) that beneﬁt themselves (Bebchuk & Fried, 2004). Hence, while principal–agency problems might be resolved by majority ownership by senior managers, reﬂecting these differing perspectives, nonmanagerial shareholders might either beneﬁt from or suffer under majority ownership by managers. A theoretical literature also examines the relationship between ownership by outsiders (especially by institutions) and managerial pay. Again there appears to be theoretical ambiguity as to outcomes (e.g., Pound, 1988). If there are strategic alliances between institutions and managers, then ﬁrm performance might suffer and be accompanied by managers receiving excessive pay. Alternatively, if outside institutions effectively monitor managers, then shareholders will beneﬁt (and presumably managers will not

120

DEREK C. JONES AND NIELS MYGIND

receive excessive compensation).7 In the case of ﬁrms in which there is majority ownership by foreigners, similar considerations might be expected to apply. Thus, to provide for improved monitoring of managers, ﬁrms owned by foreigners can be expected to be more likely to pay higher (efﬁciency) wages and to link pay with ﬁrm performance. This will contrast with ﬁrms that are owned by employee insiders who, because of closer proximity, are able to closely monitor executive performance and who will be expected to be more effective at disciplining managers than in other private ﬁrms. Finally, it is worth noting that not all who predict differences in CEO pay by type of private ownership focus on principal–agency relationships; moreover, the predictions of other theorists may sometimes differ from those who emphasize principal–agency issues. Thus, theorists who reason from a more comparative organization perspective stress that differences in ownership are also likely to be associated with differences in enterprise objectives and thus managerial rewards and the types of managers wanted by different kinds of organization. The case of employee ownership is especially pertinent. The greater likelihood of employee-owned ﬁrms having differing value systems and institutional practices such as mechanisms for employee participation in decision making (Hansman, 1996) can be expected to produce more compressed wage structures – lower levels of executive pay – than in ﬁrms with other forms of majority ownership.8 This argument is especially pertinent in those former Soviet-type economies such as Estonia where egalitarian norms were strong in the past (Mygind, 1999) and where, because of institutional inertia, such norms could have been expected to continue to exert an inﬂuence on pay structures during early transition. In addition, employee-owned ﬁrms may be expected to place a greater emphasis on goals other than proﬁts and on outputs that are more difﬁcult to measure and reward. In turn, this pursuit of different objectives produces differences in the kinds of managers required by employee owners and also in the importance of non-pecuniary rewards to managers in such ﬁrms (Frey, 1997).

PRIVATIZATION AND TRANSITION IN ESTONIA As our econometric work hinges on differences in ownership structures, it is important to consider key aspects of the legal institutional changes that occurred in Estonia during early transition. In Estonia, often privatization is discussed (e.g., European Bank for Reconstruction and Development

Majority Ownership and Chief Executive Compensation

121

[EBRD], 2000) as though all privatization had occurred in a particular way (through a state agency soliciting tenders for state ﬁrms). But in fact, the privatization process has been much more complex and the resulting range of forms of enterprise potentially is quite diverse. In Estonia, as in many other transition economies, initially the privatization process conveyed special advantages to insiders. This happened ﬁrst during the Gorbachev era in the late 1980s with the nurturing of a handful of collectively owned ﬁrms or ‘‘people’s enterprises.’’9 Subsequently, there were opportunities for employees to lease state ﬁrms, and initially, some advantages were given to employees concerning privatization of small ﬁrms. However, these advantages to employees were short lived (Kalmi, 2002). The objectives of privatization legislation changed away from limited support for employee ownership. A political climate, which altered dramatically after independence, led to the bulk of the privatization of big ﬁrms proceeding through use of a Treuhand-like privatization agency soliciting tenders for state ﬁrms. Equally, a core investor model was encouraged and foreign ownership was aggressively and fairly successfully sought (OECD, 2000). Consequently, after 1993, very few ﬁrms were privatized to employees. However, reﬂecting earlier privatization policy, it is likely that very different patterns of ownership might have been expected to have emerged within Estonia. But the situation is complicated because both theory and case studies (e.g., Kalmi, 2003) suggest that these new forms of ownership varied in stability. In particular, employee ownership is believed to have been much been less stable than other forms of ownership (e.g., Aghion & Blanchard, 1998; for Estonia, Jones & Mygind, 2005). Also, case studies of ﬁrms privatized early do not provide any strong evidence of any selection bias insofar as employees were systematically able to buy ﬁrms that were better performing than the average ﬁrm (Mygind & Pedersen, 1996). Finally, when examining the determinants of executive pay in Estonia, it is useful to take into account a number of features of transition in that country. According to many macro-economic indicators, early systemic performance in Estonia was far better than most former communist countries (Estrin et al., 2009). Second, the Estonian transition strategy included rapid price and trade liberalization and a new competition policy, and thus is reminiscent of ‘‘big bang’’ experiences elsewhere. Also, the move to privatization of large state-owned ﬁrms was quite rapid and was quite advanced during the period under study in this chapter. In these respects, the Estonian situation is very different than that prevailing in other

122

DEREK C. JONES AND NIELS MYGIND

transition and developing countries where CEO compensation has been examined.10 Thus, during early transition in Bulgaria, a focus of reform efforts was on corporatization rather than wholesale privatization. Perhaps, most important, because of the failure to deal with the problem of bad debts, the context within which Estonian managers operated was nearly always characterized by hard budget constraints – a very different situation than that which confronted managers of, for example, corporatized ﬁrms in Bulgaria, and more akin to that confronting managers of Czech ﬁrms (Eriksson, 2005).

DATA With the cooperation of the central statistical authority, annual economic and ﬁnancial data were extracted from the population of Estonian companies that existed during 1993–1997 to produce a random sample of 666 ﬁrms. This rich panel contains standard economic data, including proﬁts, sales, assets, and employment.11 However, because this data set has limited information of enterprise ownership and CEO characteristics, it was necessary for the authors to administer two special surveys and then to merge the resulting data sets. One survey focused on the distribution of ownership and aimed to collect annual data during 1993–1997 from all ﬁrms in this large panel. Reﬂecting inﬂuential literature (e.g., Aghion & Blanchard, 1998), the focus was on ownership by insiders (gathered separately for managers and employees) and non-state outsiders (split into foreigners and domestic owners) as well as ownership by the state.12 By selecting a large sample, we expect to have representation of ﬁrms in which these different owners assume varying importance and including some ﬁrms that were never privatized. The response rate for this ‘‘ownership survey’’ was very high – more than 95%. The other survey collected data for a random subsample of top executives in 220 of these ﬁrms, including information on their compensation packages.13 As CEO compensation data are not publically available in Estonia, we must rely on survey data to obtain this information. At the same time, several factors including the small size of the Estonian economy and traditions of openness concerning managers’ pay packages mean that we are conﬁdent that our data on CEO pay are reasonably accurate. Indeed in several cases, we were able to cross-check the self-reported data (through informal channels), and in all of these instances, we found no major discrepancies between the numbers reported using the different sources.

Majority Ownership and Chief Executive Compensation

123

In sum, we have no reason to believe that there are major measurement problems with any of our key variables. These data enable not only estimation of diverse speciﬁcations but also construction of measures of key variables. Concerning ownership, most studies of executive compensation in transition economies have been able to attempt to classify ﬁrms according to whether ﬁrms are state or privately owned. And even though there has been a huge amount of literature that investigates the impact of different forms of privatization upon economic performance (reviewed in Djankov & Murrell, 2002; Estrin et al., 2009), usually they are able to construct measures only of which group is the largest or the dominant shareholder (Frydman, Gray, Hessel, & Rapaczynski, 1999). However, for several reasons, this procedure does not necessarily produce the preferred typology of ownership forms. For example, dispersed shareholdings within a category may lead to limited cohesiveness by the dominant ownership group, which may account for as little as 25% of the total voting stock. However, in most cases in Estonia, the available data mean that we are able to classify ﬁrms based on the analytically preferable method of majority ownership. In Table 1, we are able to classify 214 (of 220) ﬁrms according to their majority ownership status in all years during 1993–1997, although here we report these data only for the ﬁrst and last years. One important fact concerning the nature of ownership among Estonian ﬁrms is that, during this period, there was enormous dispersion in ownership types. Although there have been instances in other economies in which some of these types of majority ownership have assumed prominence, it is most unusual, even

Table 1.

Distribution of Majority Ownership. 1993

1997

State Private Foreign Domestic Manager Employee No majority

49 165 30 53 39 43 0

11 199 42 76 59 22 4

Total

214

214

Note: Entries refer to the number of ﬁrms for which that type of owner was in the majority in that year.

124

DEREK C. JONES AND NIELS MYGIND

among transitional economies, to ﬁnd such high proportions for some of these types of ﬁrms. Thus, by 1997, in more than 27% of ﬁrms, managers had assumed majority ownership, about double the highest rate for other transitional economies for which data are available then.14 Moreover, in most other transitional economies, there tends to be a particular pattern of majority ownership that tend to emerge after privatization – for example, in the Czech Republic ﬁrms that were majority-owned tended to be mainly owned by investment funds. But in Estonia, Table 1 indicates a remarkable tendency for several types of majority ownership to coexist and instances in which there was not a majority ownership group are very rare.15 Table 1 also indicates that dramatic changes in ownership occurred during the period.16 In Table 2, we present descriptive information on key economic variables.17 By examining the entries for the whole sample, we see that, on average, during 1993–1997, real monthly CEO compensation fell very slightly, as did size when measured by sales. However, when measured by employment and assets, size fell much more sharply. For the whole sample, all performance measures – labor productivity, return on assets, and proﬁts – deteriorated sharply.18 Another purpose of Table 2 is to present descriptive information according to type of majority ownership. As majority ownership status changes in many ﬁrms during the period, we classify ﬁrms according to their ownership status at the end of the period (in 1997) and then make comparisons with how this group was doing at the start (in 1993). Thus, the entries for majority state-owned ﬁrms (in both 1993 and 1997) are obtained for those 11 ﬁrms that were majority state-owned in 1997. Table 2 reveals great differences in key variables when ﬁrms are classiﬁed according to majority ownership and that often these vary by ownership type. Firms that remained state-owned throughout the period continued to be much larger than other ﬁrms, although their rate of shrinkage was noticeably faster than for other ﬁrms.19 Unlike other ﬁrms, they also experienced rapidly increasing losses during 1993–1997. CEO pay in state ﬁrms suffered enormously too while performance indicators displayed an uneven record – for example, compare constant labor productivity and declining margins. In private ﬁrms, pay differences between managers and the average worker (RATIO) typically were quite modest – about 3:1. Interestingly, during 1993–1997, this ratio fell in foreign-owned ﬁrms (as well as in employee-owned ﬁrms), although it widened somewhat in other private ﬁrms. 20 However, the level of CEO pay was much higher in 1997 in ﬁrms majority-owned by foreigners – more than three times as high as in ﬁrms in which employees were majority owners. Unlike ﬁrms owned by mangers or

4,244 4,211 2.6 3.1 138 97 13,235 9,030 13,891 12,425 136 71 0.037 0.028 0.042 0.041 169 128

6,344 5,298 3.4 3.2 243 160 38,147 26,580 36,969 28,811 5,349 2,093 0.399 0.247 0.75 0.412 561 198

6,600 4,490 3.3 3.0 571 210 62,111 26,588 64,637 34,404 2917 4228 0.0849 .411 0.0425 0.564 132 132

Mean 3,578 1,420 2.3 1.03 681 288 95,034 37,827 74,369 72,882 2,851 7,845 0.134 0.386 0.0846 0.635 104 188

SD 7,153 6,408 3.5 2.9 89 89 26,836 23,902 22,253 29,225 1234 663 0.083 0.07 0.296 0.004 529 312

Mean 1,2710 7,444 6.2 1.8 156 140 68,673 52,418 66,194 50,091 7,798 2,497 0.88 0.177 1.6 0.134 1,211 355

SD

Foreign-Owned

3,961 4,054 2.7 3.5 157 122 10,543 7,005 12,893 9,478 1,064 44 0.067 0.013 0.029 0.1 90 87

Mean 3,853 3,783 3 3.4 200 155 18,385 14,884 22,755 13,815 6,217 1,388 0.158 0.258 0.415 0.62 131 110

SD

Domestic-Owned

3,184 3,754 2 3 102 71 5,680 2,780 7,701 5,645 135 143 0.04 0.08 0.005 0.035 87 87

Mean 2,305 5,920 1 4.3 251 183 14,920 6,374 17,124 12,040 3,092 348 0.21 0.24 0.146 0.122 83 85

SD

Manager-Owned

2,379 1,795 2 1.8 88 52 2,679 1,841 3,811 3,946 214 146 0.107 0.07 0.032 0.036 51 67

Mean

1,827 1,055 0.75 0.51 89 47 3,280 2,150 6,756 4,834 360 339 0.14 0.092 0.098 0.055 68 71

SD

Employee-Owned

Notes: All ﬁgures, including those for 1993, are for ﬁrms classiﬁed by majority ownership in 1997. As shown in Table 1, the available data enable 214 ﬁrms to be classiﬁed by majority ownership, of which, for example, 11 were majority state-owned in 1997. All variables are deﬁned in the appendix; all value variables are in 1993 kroons.

CEOPAY93 CEOPAY97 RATIO93 RATIO 97 EMPLOY93 EMPLOY97 ASSET93 ASSET97 SALES93 SALES97 PROFIT93 PROFIT97 ROA93 ROA97 MARGIN93 MARGIN97 PROD93 PROD97

SD

State-Owned

Descriptive Statistics by Type of Majority Ownership: Means (Standard Deviation).

Total

Mean

Table 2.

Majority Ownership and Chief Executive Compensation 125

126

DEREK C. JONES AND NIELS MYGIND

domestic outsiders, real CEO pay in foreign-owned ﬁrms (which includes ﬁrms that were always foreign-owned as well as ﬁrms that became foreignowned) fell during 1993–1997. Performance indicators for private ﬁrms were typically stronger than in state ﬁrms although patterns varied by majority ownership type. Thus, for ﬁrms that were majority foreign-owned at the end of the period, sharp gains in real sales and in proﬁts were recorded, although labor productivity declined considerably. Although productivity and sales grew in employee-owned ﬁrms, proﬁts fell. In ﬁrms owned by managers, proﬁts, margins, and the return on assets all improved although sales fell and productivity stagnated. Typically foreign-owned ﬁrms experienced greater rates of contraction (especially in employment) than did other privately owned ﬁrms.

THE RELATION BETWEEN EXECUTIVE COMPENSATION AND FIRM PERFORMANCE IN TRANSITION To study the determinants of the level of CEO pay, in the ﬁrst set of regressions, we augment a standard chief executive compensation equation21 by a dummy variable indicating whether, during the period 1993–1997, the ﬁrm remains state-owned. That is, ln Payit ¼ b lnðSIZEit Þ þ ðPERFORMANCEit Þ þ ðSTATEit Þ þ 8i þ Wt þ uit

(1) where Payit is chief executive pay of ﬁrm i in year t; SIZEit is size of ﬁrm i in year t; PERFORMANCEit is standard ﬁrm performance measures such as various accounting proﬁtability measures of ﬁrm i in year t; STATEit ¼ 1 in years in which a ﬁrm is state-owned, 0 otherwise; ’i is ﬁrm-speciﬁc ﬁxed effects; and Wt is year effects. The disturbance term, uit, is distributed NID(0, F2). To examine hypotheses on the impact of the type of private ownership, in the second set of regressions, we estimate Eq. (1)* where we replace STATEit in Eq. (1) by a set of four majority (private) ownership dummies (where the base case is majority state ownership).22 PERFORMANCE and SIZE have been included in prior empirical studies of executive compensation in advanced market and transitional economies. In the western literature, the application of principal–agent theory to the design of executive contracts in general predicts a positive correlation between managerial pay and some observable measure of ﬁrm

Majority Ownership and Chief Executive Compensation

127

performance (which eventually translates into improved well-being for shareholders).23 To adequately measure PERFORMANCE, the debate has usually centered on the respective merits of measures of stock market returns compared to various accounting measures such as ROA (the return on assets). However, in a context of embryonic capital and stock markets (as in Estonia during this period), this debate is moot. Moreover, many have argued (e.g., Earle & Estrin, 1998) that the key performance measure may be labor productivity. Thus, in our empirical work, as well as two accounting measures, ROA and MARGIN (gross proﬁt/sales), we also consider labor productivity as an alternative ﬁrm performance measure. SIZE is included in western studies because theory stresses the importance of factors such as spans of control in determining CEO pay. For transition economies, being a chief executive of a ﬁrm with many employees under communism often translated into more political power and thus an improved ability to obtain higher pay. To see if this force persists, our data allow us to use three alternative SIZE measures: (i) EMPLOY (number of workers); (ii) SALES (in 1993 real kroons); and (iii) ASSET (total assets in real 1993 kroons). In all estimates, we use a two-way ﬁxed effects model. Year dummy variables (Wt) capture technological change and other shocks common to all ﬁrms as well as possible measurement errors of inﬂation. Firm-speciﬁc effects (’i) are included to capture time invariant ﬁrm-speciﬁc factors that may affect chief executive pay.24 We estimated nine speciﬁcations of Eq. (1) depending on the selection of the size and performance measures.25 Importantly, for all speciﬁcations reported in Table 3, F-tests refute the joint exclusion of year dummy variables and ﬁrm-speciﬁc ﬁxed effects at the 1% level. The ﬁrst three columns of Table 3 summarize the results when employment is used as a size measure, whereas columns 4–6 report ﬁndings using sales as a measure of size and in the last three columns assets are used to proxy size. For each size measure, results are reported when the three different performance measures are used (for example, ROA in columns 2, 5, and 8). Finally, all models include a dummy variable for whether or not the ﬁrm was state-owned in that year. We begin by examining the impact of ownership on CEO pay. From Table 3, we see that the ownership coefﬁcients are statistically signiﬁcant (10% or better) in six cases and close to being statistically signiﬁcant in the remaining cases. This is reasonably strong evidence that whether a ﬁrm is privately or state-owned matters much in the determination of CEO pay, and CEOs in state ﬁrms earn higher pay then their private counterparts,

0.105 (1.688)* 0.08 922

0.03786 (1.688)*

7.7 (82.7) 0.04419 (1.888)*

I

0.1059 (1.679)* 0.10 906

0.07646 (1.722)*

7.7 (81.6) 0.04457 (1.887)*

II

0.0402 (1.537) 0.1022 (1.628) 0.23 924

7.5 (49.114) 0.056200 (1.53)

III

Statistically signiﬁcant at 10% level. Statistically signiﬁcant at 5% level.

0.12559 (2.2)** 0.23 955

0.0031 (0.205)**

0.03943 (2.136)**

7.60 (50.164)

IV

0.13607 (2.392)** 0.20 915

0.063760 (1.449)

0.031170 (1.736)*

7.63 (51.767)

V

0.01588 (0.489) 0.10221 (1.628) 0.23 924

0.056093 (2.413)**

7.5 (49.109)

VI

0.11612 (2.043)** 0.30 915

0.08106 (3.111)** 0.00786 (0.571)

7.28 (37.34)

VII

0.1186 (2.098)** 0.30 919

0.07017 (1.638)

0.080888 (3.118)**

7.27 (37.49)

VIII

The Determinants of the Level of CEO Compensation in State and Private Firms.

Notes: Absolute values of t-statistics are in parentheses.

R2 Sample size

STATE

In(PROD)

ROA

MARGIN

In(ASSET)

In(SALES)

In(EMPLOY)

Intercept

Independent Variable

Table 3.

0.022212 (0.752) 0.09929 (1.575) 0.33 904

0.07987 (2.804)**

7.19 (34.15)

IX

128 DEREK C. JONES AND NIELS MYGIND

Majority Ownership and Chief Executive Compensation

129

even after controlling for size, performance, and other time invariant unobservables. These positive and signiﬁcant estimates suggest that, after controlling for ﬁrm performance, the average CEO working for a stateowned ﬁrms receives additional rents of 10–13.6% as compared to his/her counterpart in privatized ﬁrms. Furthermore, the comparative magnitude of these ownership coefﬁcients are quite large – for example, a doubling of employment would be expected to produce only about a 5% increase in executive pay, which is far less than the effect of an ownership change. Hence, the rent earned by state ﬁrm CEOs seems to be signiﬁcant not only statistically but also economically. Put another way, the positive and signiﬁcant estimates on the coefﬁcient of STATE might be interpreted as indicating the presence of ﬁnancial discipline, which privatization has brought to CEO compensation, at least in privatized ﬁrms during early transition.26 No matter which measure of SIZE is used, evidence is found of a positive relationship, and eight of nine SIZE coefﬁcients are statistically signiﬁcant. Although results are not sensitive to the choice of the PERFORMANCE measure, the estimated pay elasticities of size are quite small and in the range of 0.03–0.08. For example, as sales increase by 10%, CEO pay increases by 0.3–0.5%; for assets, the comparable effect is 0.8%. These elasticities are much smaller than those obtained in other studies. For example, Rosen in reviewing various western studies on the estimated elasticity of pay with respect to scale ﬁnds a typical value of 0.25, whereas Jones and Kato (1997) report elasticities of size of 0.2–0.4 for Bulgaria. Finally, there is mixed support for the existence of a relationship between CEO pay and PERFORMANCE. For productivity (ln PROD), no statistically signiﬁcant relationship is ever found. However, when ROA is used to proxy performance, as hypothesized, the relationship is always found to be positive, and in one case, it is also statistically signiﬁcant. Also, in two of three cases, evidence is found of a positive and statistically signiﬁcant relationship between PERFORMANCE (measured by MARGIN) and CEO compensation. Again, these elasticities are smaller than those obtained in other studies. Thus, Rosen ﬁnds that the estimated sensitivity of pay to accounting measures are in the 1.0–1.2 range, about twice the size of the estimated semi-elasticities for Estonia for ROA. However, the ﬁnding of statistically signiﬁcant relationships between PERFORMANCE (measured by ROA or MARGIN) contrasts with the case of Bulgaria where the only signiﬁcant link was when productivity was used. This ﬁnding of a weak link between pay and performance is not unexpected. It suggests that conditions that will better facilitate managers

130

DEREK C. JONES AND NIELS MYGIND

being able to give improved attention to comparative business performance had not yet been provided been during early transition. In Table 4, we report our estimates of Eq. (1)* (when we explicitly examine the impact of the particular form of private ownership). In light of F-tests on the equality of the coefﬁcients on the ownership dummy variables, the speciﬁcations reported in Table 4 are preferred to the corresponding speciﬁcations reported in Table 3.27 Importantly, in all nine estimates, separate F-tests refute the joint exclusion of year dummy variables, the vector of private ownership dummies, and the ﬁrm-speciﬁc ﬁxed effects at the 1% level. First, we discuss ﬁndings for different forms of individual ownership. The most consistent evidence emerges when employees are majority owners. As hypothesized, relative to ﬁrms in which the state is the majority owner, ﬁrms with majority employee ownership pay their managers less, other things equal. Employee ownership coefﬁcients are statistically signiﬁcant for eight (of nine) speciﬁcations, and the size of the ownership effects is remarkably consistent across all speciﬁcations, averaging about 15%. This ﬁnding provides strong evidence for the hypothesis that, because of their close proximity, employee insiders monitor executive performance and CEO pay and thus avoid some agency costs. However, the evidence is also consistent with the view that, compared to other ﬁrms, employee-owned ﬁrms place a greater emphasis on goals other than proﬁts (such as income equality) and on outputs that are more difﬁcult to measure and reward. In turn, this pursuit of different objectives produces differences in the kinds of managers required by employee owners and in the importance of non-pecuniary rewards to managers in such ﬁrms.28 More surprisingly, perhaps, CEOs in ﬁrms in which domestic outsiders are the majority earn less than counterparts in state-owned ﬁrms. This relationship is strongly statistically signiﬁcant in four of nine speciﬁcations and quite close to being statistically signiﬁcant in remaining speciﬁcations. The effect ranges from 9 to 12% depending on speciﬁcation. This evidence is consistent with the hypothesis that external owners are monitoring managers and helping to ensure that CEOs do not receive excessive compensation. As hypothesized, ﬁrms in which ownership lies overseas always pay their managers more although these effects are never statistically signiﬁcantly different than zero (although t-statistics are usually quite a bit higher than one). Thus, these estimates, at best, provide very weak evidence in support of managers of foreign-owned ﬁrms receiving efﬁciency wages. Finally, once controls for size, performance, and time invariant ﬁrmspeciﬁc factors have been introduced, it appears that the level of CEO pay in

0.119554 (1.312) 0.099858 (1.588) 0.01417 (0.202) 0.1491 (1.888)* 0.18 922

0.03329 (1.45)

7.759 (75.442) 0.047599 (2.028)**

I

0.1449769 (1.57) 0.09802 (1.554) 0.019114 (0.272) 0.15328 (1.943)* 0.2 906

0.07027 (1.585)

7.748 (74.741) 0.04749 (2.005)**

II

0.03618 (1.381) 0.1162 (1.274) 0.09789 (1.559) 0.00973 (0.139) 0.14826 (1.879)* 0.28 924

7.56 (47.8) 0.05830 (2.5)**

III

0.08462 (0.974) 0.11731 (2.018)** 0.025988 (0.388) 0.16008 (2.109)** 0.26 955

0.00602 (0.40)

0.038311 (2.058)**

7.65 (49.21)

IV

0.102001 (1.161) 0.12096 (2.085)** 0.03752 (0.563) 0.16809 (2.225)** 0.22 915

0.057650 (1.309)

0.028470 (1.575)

7.727 (51.29)

V

0.02205 (0.679) 0.11625 (1.275) 0.09789 (1.559) 0.009704 (0.138) 0.14822 (1.879)* 0.28 924

0.05822 (2.495)**

7.6 (47.8)

VI

0.09619 (1.1) 0.1042 (1.799)* 0.01603 (0.241) 0.16369 (2.176)** 0.35 915

0.08430 (3.202)** 0.00458 (0.334)

7.32 (36.27)

VII

Dependent Variable: In (CEOPAY) (monthly pay in 1993 Kroons)

0.06951 (1.092) 0.10661 (1.844)* 0.021705 (0.327) 0.16749 (2.234)** 0.35 919

0.062475 (1.459)

0.08421 (3.213)**

7.32 (36.4)

VIII

IX

0.01508 (0.512) 0.12945 (1.403) 0.08755 (1.392) 0.00147 (0.021) 0.149807 (1.2) 0.36 904

0.08567 (2.99)**

7.23 (33.37)

The Determinants of the Level of CEO Compensation in Firms with Different Forms of Private Ownership.

Notes: Absolute values of t-statistics are in parentheses. The inclusion of all ownership dummies in all speciﬁcations is signiﬁcant at the 5% level. * Statistically signiﬁcant at 10% level. ** Statistically signiﬁcant at 5% level.

R2 Sample Size

EMPLOYEE

MANAGER

DOMESTIC

FOREIGN

In(PROD)

ROA

MARGIN

In(ASSET)

In(SALES)

In(EMPLOY)

Intercept

Independent Variable

Table 4. Majority Ownership and Chief Executive Compensation 131

132

DEREK C. JONES AND NIELS MYGIND

ﬁrms owned by managers is no different than in ﬁrms owned by the state. All managerial ownership coefﬁcients are quite tiny and strongly insignificant. As such, this evidence of the absence of excessive compensation for CEOs in managerially owned ﬁrms provides strong evidence for the view that managerial ownership, even at very high levels, can convey beneﬁts to ﬁrms and minority shareholders.29 As in the previous estimates reported in Table 3, again we ﬁnd that most measures of SIZE in Table 4 have positive and statistically signiﬁcant effects on the level of CEO pay, other things equal. The sizes of the coefﬁcients on the three proxies for SIZE are essentially unchanged from the previous estimates. However, and unlike ﬁndings reported in Table 3, in these estimates, we ﬁnd that PERFORMANCE is never statistically signiﬁcant.30 Thus, this evidence strongly suggests that the conditions that will better facilitate managers being able to give close attention to relative business performance had not yet been developed during early transition.

SUMMARY AND IMPLICATIONS By using unusual panel survey data for Estonian ﬁrms with matching information for chief executives, we study the determinants of the level of CEO compensation during early transition. As privatization led to the emergence of ﬁrms with sharply differing majority ownership conﬁgurations, we were able to examine the potential role of large differences in enterprise ownership as well as traditional forces, such as size and performance. Findings based on ﬁxed effects models indicate that CEOs in state-owned ﬁrms receive about 10–12% more pay than CEOs in private ﬁrms, ceteris paribus, which is quite sizeable. Moreover, we ﬁnd that an economically signiﬁcant determinant of CEO pay is ownership. The most consistent ﬁndings are for ﬁrms in which employees are the majority owners; relative to ﬁrms in which the state is the majority owner, ﬁrms with majority employee ownership pay their managers about 15% less, other things equal. CEO pay is also found to be statistically lower (by about 10%) in ﬁrms owned by outsiders who are not foreigners. But in ﬁrms owned by foreigners or managers, no statistically signiﬁcantly differences are found (compared to CEO pay in state ﬁrms). However, we are aware that our ﬁndings are potentially limited by some limitations of the data. In particular, we are unable to include individual CEO ﬁxed effects. It is certainly plausible that managers could change during this period of early transition (and that new compensation contracts could be formed during transition). At the same time, we note that often

Majority Ownership and Chief Executive Compensation

133

when managerial turnover has been included in managerial performance– pay regressions in other transition economies, results do not seem to be strongly affected (e.g., Eriksson, 2005). Also we are aware that there are potential econometric issues surrounding ownership and endogeneity. It is possible that highly paid managers inﬂuence ownership type and that complications may arise from non-salary types of compensation or actual ownership holdings of top managers. Unfortunately, the available data do not enable us to address such matters in a satisfactory way (e.g., there do not appear to be suitable instruments to enable instrumental variable estimates). Subject to these potential limitations, the ﬁnding that ownership, including the particular form of ownership, has important effects for the level of CEO pay potentially has broad implications. Previous research on the determinants of executive pay has established that institutional arrangements do have economic consequences. For example, concerning ownership by top managers, Holderness and Sheehan (1989) provide examples of excessive compensation (although Ang, Hauser, & Lauterbach, 1997 do not uncover similar evidence). Also, there is evidence that different managerial labor markets exist for ﬁrms with different objective functions – for example, Roomkin and Weisbrod (1999) ﬁnd that monetary compensation is substantially higher in for-proﬁt hospitals than in nonproﬁts. But our ﬁndings are for a case where extraordinary differences in ownership structures and institutional conditions existed and thus enable much more general tests of such propositions to be undertaken and more general conclusions to be drawn. Thus, concerning ownership, a large body of work (e.g., Estrin et al., 2009) and including the particular case of Estonia (Jones & Mygind, 2002) points to the beneﬁts of privatization for improved business performance. The evidence presented in this chapter clearly reinforces the conclusion that there are beneﬁts of privatization. It appears that the arrangements that existed in privately owned ﬁrms in Estonia enabled all types of privately owned ﬁrms to do a better job of monitoring CEO pay than was possible under the institutional arrangements that existed in state-owned ﬁrms and which enabled CEOs to continue to engage in rent seeking.31 Findings on the links between the form of private ownership and CEO pay provide evidence that relates to diverse hypotheses. Thus, the ﬁnding that CEO pay in ﬁrms owned by managers was not excessive (relative to state-owned ﬁrms or foreign-owned ﬁrms) provides strong evidence for the view (e.g., Jensen & Meckling, 1976) that managerial ownership, even at very high levels, can convey beneﬁts to ﬁrms and shareholders. However, in some cases, our ﬁndings are consistent with more than one hypothesis. In particular, our ﬁnding that CEO compensation is signiﬁcantly lower in

134

DEREK C. JONES AND NIELS MYGIND

employee-owned ﬁrms is consistent with the principal–agency story that employee owners are able to reduce agency costs by effectively monitoring managers. But this evidence is also consistent with the view that employeeowned ﬁrms pursue different objectives than other ﬁrms. This produces differences in the kinds of managers required by employee owners and also in the importance of non-pecuniary rewards to managers in such ﬁrms. Finally, arguably, our ﬁndings do not provide evidence that, from the perspective of CEO pay, there is an optimal form of private ownership. As such, this conclusion mirrors much of the evidence that examines links between types of private ownership and enterprise performance. Simultaneously, we investigated the role of more traditional determinants of executive pay during this interesting period of early transition. These ﬁndings indicate that CEO pay is always positively and usually statistically signiﬁcantly related to variation in size, which is measured by sales, assets, and employment. The evidence on the link between CEO compensation and performance is much more mixed. Using measures such as proﬁt margin and ROA, evidence of a positive and statistically signiﬁcant relationship is sometimes found when estimates include only a single ownership dummy (indicating whether a ﬁrm is privately owned). However, in the preferred speciﬁcations, when we introduce several ownership dummies reﬂecting differences in types of majority private ownership, no link is found between pay and productivity. Together with the strong compensation–size link and the very weak pay– performance relationships, our ﬁndings on the importance of ownership for CEO pay suggest that the ﬁnancial discipline surrounding ﬁrms in the Estonian state sector was especially weak during the period under investigation. In other words, arguably, privatization has imposed strong discipline on the level of CEO compensation in Estonia. However, compared to ﬁndings from other studies, the size and performance semielasticities typically are quite small, averaging about 0.04 and 0.06, respectively. These ﬁndings suggest that executive compensation in Estonia is set so as to provide only weak incentives for managers to pay limited attention to ﬁrm size and almost no attention to current business performance. This ﬁnding of a weak link between pay and performance has been found in other transition economies, including Bulgaria (Jones & Kato, 1996) although Groves, Hong, McMillan, and Naughton (1995) ﬁnd a strong pay–performance link in China. In accounting for the differences across countries on these points, clearly, the differences in institutional contexts matter. In addition, the failure to ﬁnd a link between productivity/ proﬁtability and pay in Bulgaria is unsurprising because, unlike in China,

Majority Ownership and Chief Executive Compensation

135

the performance of the Bulgarian economy has been so chaotic. But the existence of a very weak pay–performance link in Estonia, especially during a period when it is believed that Estonian ﬁrms operated under hard budget constraints, is more difﬁcult to explain. Finally, our ﬁnding of the strong role played by employee ownership in limiting executive pay is potentially of interest to policy makers who seek mechanism to rein in executive pay that is sometimes viewed as excessive. Although employee ownership and related forms of shared capitalism appear to be spreading around the globe, further encouragement of the growth of employee ownership would likely deliver an additional societal gain besides those to workers and ﬁrms highlighted in recent research (e.g., Kruse, Freeman, & Blasi, 2010).

APPENDIX. DEFINITIONS OF VARIABLES CEOPAY ¼ Real monthly salary of CEO (1993 kroons) RATIO ¼ (CEOPAY/average monthly salary of all employees) SIZE ¼ A measure of size EMPLOY ¼ Employment ASSET ¼ Real total assets (thousands 1993 kroons) SALES ¼ Real sales (thousands 1993 kroons) PERFORMANCE ¼ A measure of enterprise performance PROFIT ¼ Real proﬁts (thousands 1993 kroons) ROA ¼ Rate of return on assets ¼ PROFIT/ASSET MARGIN ¼ Real proﬁt margin (thousands 1993 kroons) PROD ¼ Real sales per worker (thousands 1993 kroons) STATE ¼ 1 if majority owner is state, 0 is otherwise or PRIVATE where private owners own a majority of the voting equity FOREIGN ¼ 1 if majority owners are foreigners, 0 otherwise DOMESTIC ¼ 1 if majority owners are domestic outsiders (e.g., ﬁrms registered in Estonia, individual Estonians who do not work at the ﬁrm), 0 otherwise EMPLOYEE ¼ 1 if majority owners are nonmanagerial employees, 0 otherwise MANAGER ¼ 1 if majority owners are managers, 0 otherwise NOMAJORITY Are ﬁrms in which there was no majority owner Notes: A sufﬁx of 93 indicates that variable is for 1993; a sufﬁx of 1997 signiﬁes the observation is for 1997. All value variables are in real 1993 kroons and deﬂators are taken from EBRD (2000).

136

DEREK C. JONES AND NIELS MYGIND

NOTES 1. Morck et al. (1988) ﬁnd a nonlinear relationship between the extent of managerial ownership (reﬂecting managerial entrenchment) and Tobin’s Q. Related issues are also examined by Joskow, Rose, and Shepard (1993). 2. However, there is an extensive literature that examines diverse issues for ﬁrms that are majority-owned by nonmanagerial employees. For one example, see Pencavel and Craig (1994); for a review of such studies, none of which examine the determinants of CEO pay, see Bonin, Jones, andPutterman (1993) and Dow (2003). 3. By contrast, majority ownership structures of enterprises in organizations in western economies are relatively homogeneous. Thus, when ownership by nonmanagerial employees occurs, this seldom exceeds 5% of voting equity (Blasi & Kruse, 1991). Although ownership by top managers is more common, instances of managers owning the bulk of shares in a western ﬁrm are rare. For example, in a survey of US corporations, Denis, Denis, and Sarin (1997) ﬁnd that in about 5% of cases (282/5,545 man-years), ofﬁcers and directors had majority ownership in that ﬁrm. 4. The most comprehensive investigation of diverse aspects of a managerial labor market for a transition economy is Eriksson (2005). For general surveys of literature on managerial labor markets in former Soviet economies as well as emerging economies such as China, see Djankov and Murrell (2002) and Estrin et al. (2009). As pointed out in those surveys, our understanding of these aspects of managerial labor markets in those countries is conﬁned to a handful of empirical studies that cover a limited range of countries. 5. One exception is Roomkin and Weisbrod (1999) who examine CEO pay and the comparative behavior of for proﬁt and nonproﬁt ﬁrms. 6. Case study evidence also suggests that the managerial labor market is rather different in state ﬁrms than in private ﬁrms, with state ﬁrms typically having less turnover and relatively more internal promotion (see Mygind & Pedersen, 1996). 7. However, if ﬁrms are largely owned by outsiders but ﬁrms are part of groups, then institutions connected to the group may play a monitoring role. See Kato and Rockel (1992) for a discussion of the Japanese case. 8. Also, more compressed internal wage differences have been found in actual employee ﬁrms, including Mondragon. See, for example, the evidence in Arando, Freundlich, Gago, Jones, and Kato (2011). 9. In these ﬁrms, ownership was to be shared equally by all employees (Mygind, 1999). 10. For China, see Kato and Long (2006), for the Czech Republic see Claessens and Djankov (1999) and Eriksson (2005), and for Bulgaria see Jones and Kato (1996). 11. These ofﬁcial Estonian data have tended to be regarded as quite reliable certainly when compared with data for many other transitional countries (see OECD, 2000). 12. It was impossible to gather information on the distribution of ownership within ownership classes. However, theoretical and empirical literature for transition economies has stressed the importance of categories of ownership for enterprise behavior (e.g., Aghion & Blanchard, 1998; Jones, 2004; for the Estonian case, Jones & Mygind, 2002). 13. Comparisons of means for key economic variables in this managerial survey of 220 ﬁrms and for the larger ownership survey of 666 do not reveal any statistically

Majority Ownership and Chief Executive Compensation

137

signiﬁcant differences. Thus, there is no reason to believe that this smaller sample is not representative of the population of Estonian ﬁrms that existed during 1993–1997. 14. Although data for Russia are somewhat fragile, in 1997, it appears that in about 15%, managers constituted the dominant group (see Buck, Filatotchev, Wright, & Zhukov, 1999). 15. To be precise, there were only 10 observations of ‘‘no-majority’’ during the whole period – fewer than 1% of the total sample of 1,070. Hence, we restrict our empirical work to those 1,060 observations for ﬁrms in which there was a majority owner. 16. The nature and determinants of ownership changes in Estonian ﬁrms during a period that partially overlaps with the period examined in this chapter is investigated in greater length by the authors elsewhere (see Jones andMygind, 1999). 17. Typically, the number of observations on economic variables in each ownership category is usually very close to the numbers reported in Table 1 for 1997 – that is, a total of 214 observations, with 11 for ﬁrms that were majority stateowned and so on. However, the panel is unbalanced with the number of observations for which data on all variables are available ranging from 909 to 955. There were signiﬁcantly more cases of missing values for the earlier years with asset data being especially likely to be missing. 18. Such patterns were not unusual during early transition in many economies. See, for example, Fischer, Ratna, and Vegh (1996). 19. However, in part reﬂecting the small size of the Estonian economy, Estonian ﬁrms have always been much smaller than their counterparts in, for example, the former USSR. In turn, presumably this translated into comparatively less overmanning at the start of transition. 20. As such, this ratio of CEO/average worker pay is rather lower than what has been reported for western countries although comparable to what has been observed for some transition economies. Thus, Kato and Rockel (1992) report comparable ratios for Japan of 13 and for the United States of 32. In 1993, the comparable ratio in Bulgaria was 2.9 (Jones & Kato, 1996). 21. To facilitate comparisons of the size of key parameters, our empirical strategy mirrors that used in early studies as reviewed in Murphy (1999). Unfortunately, we are unable to include managerial ﬁxed effects – we do not have information on turnover. 22. Although there are potential econometric pitfalls surrounding issues such as endogeneity and ownership type, unfortunately we do have suitable instruments to begin to address these matters. 23. Executive compensation studies have been mainly undertaken for developed economies and do not measure performance relative to an industry mean, even though such data often are publically available in the west. For transition economies, the use of performance measures relative to industry means was difﬁcult to implement during early transition because such data were not publically available and the relatively few ﬁrms in most Estonia industries would have meant that even if such a procedure were feasible it would have been especially sensitive to outliers. 24. Unfortunately, we do not have information on individual CEO attributes such as experience and education and other components of compensation. 25. As indicated earlier, our estimates are for unbalanced panels. If the identical speciﬁcations are estimated using a balanced panel for 169 ﬁrms, the ﬁndings are essentially unaltered from those reported in the text.

138

DEREK C. JONES AND NIELS MYGIND

26. Alternatively these payments might be a compensating differential – running a state-owned ﬁrm might be viewed as either more difﬁcult than a private ﬁrm or less prestigious. 27. Nevertheless we report the results contained in Table 3 in view of our interest in the hypotheses concerning pay determination for CEOs in private versus state ﬁrms. 28. This is consistent with ﬁndings from case studies, as discussed by Mygind (1999). 29. Note also that CEO pay in ﬁrms in which managers are majority owners is also about 25% less than comparable pay in foreign-owned ﬁrms and only about 13% more than in forms in which employees are majority owners. 30. In part, this follows because performance is related to ownership. Ownership is also potentially endogenous (although our institutional discussion suggests that at least for employee-owned ﬁrms, their selection may usually have been arbitrary). In any event, unfortunately, the existing data do not contain any suitable instruments for ownership. 31. The case study evidence suggests that the average size of the board of directors is smaller in Estonian ﬁrms (compared to state ﬁrms) and that state ﬁrms continue to have some directors who are chosen primarily for political reasons.

ACKNOWLEDGMENTS The authors acknowledge support from the National Council on Eurasian and Eastern European Research, ACE Phare, and the Danish Research Council for Social Sciences and. The chapter has beneﬁtted from presentations at seminars at Hitotsubashi University and Copenhagen Business School and comments from Jeffrey Pliskin.

REFERENCES Aghion, P., & Blanchard, O. (1998). On privatization methods in Eastern Europe and their implications. Economics of Transition, 6(1), 87–99. Aghion, P., Blanchard, O., & Burgess, S. (1994). The behavior of state ﬁrms in Eastern Europe, pre-privatization. European Economic Review, 38, 1327–1349. Ang, J., Hauser, S., & Lauterbach, B. (1997). Top executive compensation under alternative ownership and governance structures. Advances in Financial Economics, 3, 1–32. Arando, S., Freundlich, F., Gago, M., Jones, D. C., & Kato, T. (2011). Assessing Mondragon: Stability and institutional adaptation in the face of globalization. In: E. C. Carberry (Ed.), Employee ownership and shared capitalism (Ch. 9). Ithaca, NY: Cornell University and LERA Research Volume. Aslund, A. (1999). The end of rent-seeking: The end of the post-communist transformation. In: A. N. Brown (Ed.), When is transition over? (pp. 51–68). Kalamazoo, MI: Upjohn Institute. Bebchuk, L. A., & Fried, J. M. (2004). Pay without performance: The unfulﬁlled promise of executive compensation. Cambridge, MA: Harvard University Press. Blasi, J., & Kruse, D. (1991). The new owners. New York, NY: Harper Business.

Majority Ownership and Chief Executive Compensation

139

Bonin, J. (1976). On the design of managerial incentive systems in a decentralized planning environment. American Economic Review, 66(4), 682–687. Bonin, J., Jones, D. C., & Putterman, L. (1993). Theoretical and empirical studies of producer cooperatives: Will the twain ever meet?. Journal of Economic Literature, 31(September), 1290–1320. Boycko, M., Schleifer, A., & Vishny, R. (1996). A theory of privatization. Economic Journal, 106(March), 309–319. Buck, T., Filatotchev, I., Wright, M., & Zhukov, V. (1999). Corporate governance and employee ownership in an economic crisis: Enterprise strategies in the former USSR. Journal of Comparative Economics, 27(3), 459–474. Claessens, S., & Djankov, S. (1999). Enterprise performance and management turnover in the Czech Republic. European Economic Review, 1115–1124. Demsetz, H., & Lehn, K. (1985). The structure of corporate ownership: Causes and consequences. Journal of Political Economy, 93, 1155–1177. Denis, D. J., & Denis, D. K. (1994). Majority owner-managers and organizational efﬁciency. Journal of Corporate Finance, 1, 91–118. Denis, D. J., Denis, D. K., & Sarin, A. (1997). Ownership structure and top executive turnover. Journal of Financial Economics, 45, 193–221. Djankov, S., & Murrell, P. (2002). Enterprise restructuring in transition: A quantitative survey. Journal of Economic Literature, 40(3), 739–792. Dow, G. K. (2003). Governing the ﬁrm. Cambridge, MA: Cambridge University Press. Earle, J. S., & Estrin, S. (1998). Workers’ self management in transitional economies. Advances in the Economic Analysis of Participatory and Labor-Managed Firms, 6, 3–28. Eriksson, T. (2005). Managerial pay and executive turnover in the Czech and Slovak Republics. Economics of Transition, 13(4), 659–677. Estrin, S., Hanousek, J., Koc, E., & Svejnar, J. (2009). The effects of privatization and ownership in transition economies. Journal of Economic Literature, 47(3), 699–728. European Bank for Reconstruction and Development (EBRD). (2000). Transition report for 2000. EBRD, London. Fischer, S., Ratna, S., & Vegh, C. A. (1996). Stabilization and growth in transition economies: The early experience. Journal of Economic Perspectives, 10(2), 45–66. Frey, B. (1997). On the relationship between intrinsic and extrinsic work motivation. International Journal of Industrial Organization, 15, 427–439. Frydman, R., Gray, C., Hessel, M. P., & Rapaczynski, A. (1999). Private ownership and corporate performance: Some lessons from transition economies. Quarterly Journal of Economics, 114, 1153–1191. Frydman, C., & Jenter, D. (2010). CEO compensation. NBER Working Paper no. 16585. Retrieved from www.nber.org/papers/w16585, Cambridge. MA. Groves, T., Hong, Y., McMillan, J., & Naughton, B. (1995). China’s evolving managerial labor market. Journal of Political Economy, 103(4), 873–892. Hansman, H. (1996). The ownership of enterprise. Cambridge, MA: Harvard University Press. Holderness, C. G., & Sheehan, D. P. (1989). What constrains managers who own large blocks of stock? Working Paper. Purdue University, West Lafayette, IN. Ickes, B., & Samuelson, L. (1987). Job transfers and incentives in complex organizations: Thwarting the Ratchet effect. RAND Journal of Economics, 18, 275–286.

140

DEREK C. JONES AND NIELS MYGIND

Jensen, M., & Meckling, W. (1976). Theory of the Firm: Managerial behavior, agency costs and ownership structure. Journal of Financial Economics, 3, 305–360. Jones, D. C. (2004). Ownership and participation: A review of empirical evidence for transition economies. In: V. Pe´rotin and A. Robinson (Eds.), Employee participation, ﬁrm performance and survival, Advances in the Economic Analysis of Participatory and LaborManaged Firms. (Vol. 8, pp. 171–209). Amsterdam, San Diego, CA, and Oxford: Elsevier, JAI. Jones, D. C., & Kato, T. (1996). The determinants of chief executive compensation in transitional economies: Evidence from Bulgaria. Labour Economics, 3(October), 319–336. Jones, D. C., & Mygind, N. (1999). The nature and determinants of ownership changes after privatization: Evidence from Estonia. Journal of Comparative Economics, 27, 422–441. Jones, D. C., & Mygind, N. (2002). Ownership and productive efﬁciency: Evidence from Estonia. Review of Development Economics, 6(2), 284–301. Jones, D. C., & Mygind, N. (2005). Corporate governance cycles during transition: Theory and evidence for the Baltics. In: I. Filatotchev & M. Wright (Eds.), Corporate governance life cycle (pp. 253–286). London: Edward Elgar Publishing. Joskow, P. L., Rose, N. L., & Shepard, A. (1993). Regulatory constraints on CEO compensation. Brookings Papers on Economic Activity, 1–58. Kalmi, P. (2002). Does employee ownership survive the transition? Case study evidence from Estonia. Acta Oeconomica, 52(3), 307–325. Kalmi, P. (2003). The rise and fall of employee ownership in Estonia, 1987–2001. Europe-Asia Studies, 55(8), 1213–1239. Kato, T., & Long, C. (2006). Executive compensation, ﬁrm performance, and corporate governance in China: Evidence from ﬁrms listed in the Shanghai and Shenzhen Stock Exchanges. Economic Development and Cultural Change, 54(4), 945–983. Kato, T., & Rockel, M. (1992). Experiences, credentials, and compensation in the Japanese and U.S. Managerial labor markets: Evidence from New Micro Data. Journal of the Japanese and International Economies, 6(1), 30–51. Kornai, J. (1992). The socialist system. Princeton, NJ: Princeton University Press. Kruse, D., Freeman, R. B., & Blasi, J. R. (2010). Shared capitalism at work: Employee ownership, proﬁt and gain sharing, and broad-based stock options. Chicago, IL: University of Chicago Press. Morck, R., Schleifer, A., & Vishny, R. W. (1988). Management ownership and market valuation: An empirical analysis. Journal of Financial Economics, 20, 293–315. Murphy, K. J. (1999). Executive compensation. In: O. Ashenfelter & D. Card (Eds), Handbook of labor economics, v. 3. Amsterdam: North-Holland. Mygind, N. (1999). Privatization, governance and restructuring of enterprises in the Baltics. CEES, Copenhagen Business School, Working Paper no. 28. Mygind, N., & Pedersen, P. N. (1996). Privatization and ﬁnancial participation in the Baltic Countries: Case studies. CEES Working Paper, Copenhagen Business School, March. OECD. (2000). Economic surveys: Baltic states: A regional economic assessment-A decade of Baltic transition. Paris. Pencavel, J., & Craig, J. (1994). The empirical performance of orthodox models of the ﬁrm: Conventional ﬁrms and worker co-operatives. Journal of Political Economy, 102, 718–744.

Majority Ownership and Chief Executive Compensation

141

Pound, J. (1988). Proxy contests and the efﬁciency of shareholder oversight. Journal of Financial Economics, 20, 237–265. Putterman, L. (1993). Division of labor and welfare: An introduction to economic systems. New York, NY: Oxford University Press. Roomkin, M., & Weisbrod, B. (1999). Managerial compensation and incentives in for-proﬁt and non-proﬁt hospitals. The Journal of Law, Economics, & Organization, 15(3), 750–781.

WORKER ATTITUDES TOWARD EMPLOYEE OWNERSHIP, PROFIT SHARING AND VARIABLE PAY Fidan Ana Kurtulus, Douglas Kruse and Joseph Blasi ABSTRACT Using the NBER Shared Capitalism Database comprised of over 40,000 employee surveys from 14 ﬁrms, we investigate worker attitudes toward employee ownership, proﬁt sharing, and variable pay. Speciﬁcally, our study uses detailed survey questions on preferences over proﬁt sharing, forms of employee ownership like company stock and stock option ownership, as well as preferences over variable pay in general, to explore how preferences for these different types of outputcontingent pay vary with worker risk aversion, residual control, and views of co-workers and management. Our key results show that, on average, workers want at least a part of their compensation to be performance-related, with stronger preferences for output-contingent pay schemes among workers who have lower levels of risk aversion, greater residual control over the work process, and greater trust of coworkers and management.

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 143–168 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012010

143

144

FIDAN ANA KURTULUS ET AL.

Keywords: Employee ownership; proﬁt sharing; variable pay; worker preferences; residual control; risk aversion; perceptions of co-workers and management JEL classiﬁcations: J54; J33; M52

INTRODUCTION The prevalence of employee ownership, proﬁt sharing and other performance-based pay schemes has been growing in the past several decades in the United States and other advanced economies. According to the 2006 wave of the General Social Survey, which is a nationally representative survey of individuals conducted by the National Opinion Research Center, over a third of U.S. workers are covered by proﬁt sharing, 18 percent own company stock, and 9 percent own company stock options. Coverage is similar in France, Great Britain, Italy and Japan (Del Boca, Kruse, & Pendelton, 1999; Jones & Kato, 1995). A large part of the previous research on shared capitalist pay schemes in which employees participate in the ﬁnancial performance of their place of work has focused on the effects of such programs on worker and ﬁrm outcomes like productivity, turnover, and proﬁts (Craig & Pencavel, 1992; Kruse, 2002; Kruse & Blasi, 1997; Park, Kruse & Sesil, 2004). But an important aspect that has not yet been explored is worker preferences for different participatory compensation programs, largely due to the dearth of available datasets that are conducive to the analysis of this subject. The current study sheds light on this topic by examining preferences over proﬁt sharing, forms of employee ownership like company stock and stock option ownership, as well as preferences over variable pay in general, and how these preferences depend on worker risk aversion, residual control, and perceptions of co-workers and management. Economic theory predicts that workers will be more favorable toward performance-related pay schemes when they: (i) have low levels of risk aversion, (ii) have greater control over the work process generating payouts (residual control), (iii) trust their co-workers, so that the free rider problem associated with group incentives can be overcome by a cooperative solution, and (iv) trust their managers not to exploit information asymmetries when distributing ﬁnancial payouts. We investigate the role that each of these factors play in workers’ preferences over employee ownership, proﬁt sharing and variable pay using a unique set of questions asked in the NBER Shared Capitalism Survey of more than 40,000 employees from 14 ﬁrms.

Worker Attitudes Toward Employee Ownership

145

This is a novel research area in the employee ownership literature and our ﬁndings help to understand how workers respond to different types of participatory compensation schemes. We consider both pay that is tied to overall company performance (proﬁt sharing, company stock and stock option ownership), as well as individual performance-based variable pay (individual bonuses, and commissions), and we will refer to these collectively as ﬁnancial participation. A strength of our data is that we have individual-level measures of risk aversion, which is often discussed as an important factor in ﬁnancial participation but is rarely measured. The NBER Shared Capitalism Survey additionally provides unique information on worker residual control and worker perceptions of co-workers and management which are also central to our analysis. We use this detailed information to investigate how preferences for different types of ﬁnancial participation are shaped by worker risk aversion, residual control, and views of co-workers and management. Our key results show that most workers want at least a part of their compensation to be output-contingent, with stronger preferences for performance-related pay among workers who have lower levels of risk aversion, greater residual control of the work process, and greater trust of co-workers and managers.

THEORY AND PRIOR LITERATURE Our discussion focuses on the following theoretical factors central to perceptions about shared capitalism and variable pay: risk aversion, residual control, and trust in co-workers and management. Worker preferences for employee ownership and other forms of ﬁnancial participation will reﬂect the perceived potential costs and beneﬁts of such plans. Risk aversion is viewed as a key factor in most theoretical models of pay-for-performance (Holmstrom, 1979; Shavell, 1979), since the variability of rewards can represent a signiﬁcant cost for risk-averse workers, and has indeed been found to reduce preferences for output-contingent pay in laboratory experiments (Cadsby, Song, & Tapon, 2007). Moreover, people with lower wealth and base salary will generally be more averse to ﬁnancial risk since they have less money for discretionary spending and a reduction in income or assets may force them to cut back on necessities. Attitudes toward ﬁnancial participation will also depend on the perceived potential for higher income, which will depend on worker skills and opportunities for inﬂuencing workplace performance. Workers are likely to view group-based rewards more favorably in the presence of practices such as

146

FIDAN ANA KURTULUS ET AL.

employee involvement in decisions (increasing opportunities to inﬂuence performance), training (increasing skills that contribute to performance), and job security (providing assurance that one will be able to receive the fruits of higher performance). This can be thought of in the context of the theory of residual returns and residual control, which argues that those who receive residual returns (ﬁnancial participation) should also receive residual control (power over the work process) in order to provide proper incentives and achieve value-maximizing decisions (Holmstrom & Milgrom, 1994; Jensen & Meckling, 1992; Milgrom & Roberts, 1990; Prendergast, 2002). Several empirical studies have found support for this hypothesis (Ben-Ner, Kong & Lluis, 2010; DeVaro & Kurtulus, 2010; Foss & Laursen, 2005). Moreover, employee involvement in ﬁrm decision making may create expectations or desires for sharing in the fruits of those decisions, and workers may become dispirited without some sort of ﬁnancial reward tied to the consequences of those decisions (Ben-Ner & Jones, 1995; Levine & Tyson, 1990). Preferences for rewards based on company performance are also likely to critically depend on workers’ perceptions about co-workers. The well-known free rider problem in group incentives has been modeled as a prisoners’ dilemma game, in which each participant has an individual incentive to shirk. If the game is repeated in an ongoing relationship, however, several equilibria are possible, including a cooperative equilibrium in which the participants establish a collective agreement to cooperate so that the rewards are higher for all participants (Axelrod, 1984; Fudenberg & Maskin, 1986). Workers in group incentive plans may establish and maintain a commitment to high work standards through cooperation and monitoring, which can generate higher payouts for workers than in a noncooperative setting (Weitzman & Kruse, 1990). This points to co-worker relations as a key ingredient in the effectiveness of employee ownership. Preferences for employee ownership and proﬁt sharing are likely to be low if workers think co-workers are not interested in workplace performance and there is little potential for productive cooperation under group-based rewards, and higher if they think co-workers are interested in workplace performance and can achieve the cooperative equilibrium in the prisoner’s dilemma game by working well together. Finally, preferences for ﬁnancial participation are likely to be shaped by attitudes toward management. First, workers are unlikely to favor variable pay plans if they do not trust managers to manage well so that there will be rewards to distribute to workers for their hard work. The second issue stems from informational asymmetries inherent in many incentive plans: it can be difﬁcult for workers to determine whether rewards under variable pay systems are being calculated correctly and fairly by management, whereas

Worker Attitudes Toward Employee Ownership

147

this is easier under ﬁxed wage contracts since workers know exact pay levels ex ante. One of the objections of unions to proﬁt sharing, for example, is mistrust that managers will calculate proﬁts in a way that properly rewards workers for their performance (Zalusky, 1986, 1990). If workers do not trust managers to calculate the payouts from ﬁnancial participation in a competent and honest way, they will be less interested in participating. In sum, based on theory and past research, we expect attitudes toward ﬁnancial participation to be more positive among workers who are less risk averse, have more residual control, and who trust management and think co-workers are more interested in workplace performance. There has been little published research on the topic of worker attitudes toward ﬁnancial participation. Kruse and Blasi (1999) summarize 30 public polls conducted between 1975 and 1997 with questions on general perceptions about ﬁnancial participation systems, ﬁnding that a majority of people expressed favorable views of employee ownership and its effects on workplaces; for instance, in one survey 75 percent said they would like to work for an employee-owned and employee-controlled company, as opposed to a company owned by outside investors or government, and in another survey 69 percent thought employee-owners work harder. A number of studies have examined the effect of variable pay on overall job satisfaction among British workers: Green and Heywood (2008) found that performance-related pay in general was associated with increased job satisfaction, Brown and Sessions (2003) showed that workers who participated in performance bonuses, share ownership, and proﬁt sharing were more satisﬁed with their work environment, and McCausland, Pouliakas, and Theodossiou (2005) found that the inﬂuence of performance pay increased satisfaction for the more highly paid but lowered it for the less highly paid. Drago, Estrin and Wooden (1992) found the use of individual and group bonuses to be a positive determinant of job satisfaction in a sample of Australian workers. Cornelissen, Heywood and Jirjahn (2008) showed that among German workers who received performance pay, risk aversion was inversely correlated with overall job satisfaction. Our study contributes to the literature by investigating worker preferences for various forms of ﬁnancial participation and how these preferences depend on key worker and workplace characteristics.

DATA AND VARIABLES We use the NBER Shared Capitalism Database, which consists of detailed information collected from more than 40,000 employee surveys from 14

148

FIDAN ANA KURTULUS ET AL.

ﬁrms, to explore preferences for different forms of employee ownership and variable pay. The NBER data comprise one of the largest worker-level datasets on labor practices and worker sentiment ever collected. The survey was conducted during 2002–2006 using a combination of web-based and paper survey methods, and had a high response rate, averaging 53 percent over the 14 companies. The ﬁrms participating in the survey included large multinationals with employment spanning North America, South America, Europe, and Asia, as well as smaller ﬁrms with mostly US employees. The sample included eight ﬁrms in the manufacturing industry, two hightechnology ﬁrms, and four in the service industry. Three of the fourteen companies exceeded 10,000 employees, 5 employed between 1,000 and 10,000 workers, and the remaining 6 employed fewer than 1,000 workers. All of the ﬁrms had employee ownership, proﬁt sharing, and variable pay programs, though of varying forms and degrees: 13 had individual bonus plans, 9 had workgroup-based or department-based performance bonus plans, 11 had broad-based proﬁt sharing plans, 5 had broad-based stock option plans, 8 had standard employee stock ownership plans (ESOPs), 1 had a 401(k) employee stock ownership program, 4 had employee stock purchase plans, and 3 had 401(k)’s with company stock. Most had combinations of these plans. To investigate the role that risk aversion, residual control, and perceptions of co-workers and management play in workers’ preferences for ﬁnancial participation, we make use of a unique set of questions asked in the NBER Shared Capitalism Survey capturing these concepts. We now turn to a discussion of the key variables used in our empirical analyses.

Dependent Variables We examine four variables reﬂecting worker preferences for ﬁnancial participation as dependent variables in our regression analyses. The ﬁrst one captures worker preferences for variable pay broadly and is based on a question asked in the NBER Shared Capitalism Survey indicating the percentage of pay the respondent would like to receive as variable compensation (which includes all forms of output-contingent pay that is based on individual, group, and company performance). Next we examine worker preferences for various forms of employee ownership based on questions indicating the respondent’s preference for being paid at least in part based on company performance (proﬁt sharing, company stock, or stock options), preference for the extent to which the respondent’s next pay increase

Worker Attitudes Toward Employee Ownership

149

depends on company performance, and preference for getting a portion of compensation in the form of company stock and stock options. These four dependent variables are deﬁned formally below. Worker Preferences for Financial Participation: Preference for variable pay

Proportion of pay the worker would like to receive as variable compensation (13 ﬁrms surveyed)

Preference for company-based incentives

Dummy variable equaling 1 if the worker prefers that he or she be paid in part with a variable amount dependent on company performance, through proﬁt sharing, company stock, or stock options; 0 if all ﬁxed wage or salary, with no proﬁt sharing, company stock, or stock options (13 ﬁrms surveyed)

Preference for company-based incentives in next pay increase

Worker’s preference that his or her next pay increase come in the form of 1 ¼ All ﬁxed wages, with no proﬁt sharing, company stock, or stock options; 2 ¼ Split between ﬁxed wages and proﬁt sharing, company stock, or stock options; 3 ¼ All in the form of proﬁt sharing, company stock, or stock options (5 ﬁrms surveyed)

Preference for stock over cash incentives

Worker’s preference for getting some of his or her compensation from company stock and stock options as opposed to a cash incentive plan on a 1 (cash incentive plan) to 5 (company stock and stock options) scale (3 ﬁrms surveyed)

Key Independent Variables The key independent variables in our analysis are those that capture worker risk aversion, residual control, and perceptions of co-workers and management.

150

FIDAN ANA KURTULUS ET AL.

One of the unique features of the NBER Survey is the presence of information on individual-level risk aversion, which plays a central role in theoretical models of the employee–employer relationship, but is rarely available in existing datasets. Our primary measure of the extent to which the worker is averse to risk is based on the NBER Survey question ‘‘Some people like to take risks and others dislike taking risks. Where would you place yourself on a scale of how much you like or dislike taking risks, where 0 is hating to take any kind of risk and 10 is loving to take risks?’’, from which we deﬁne variable Risk Averse such that values greater than or equal to 7 on this scale correspond to ‘‘low risk aversion,’’ greater than 3 and less than 7 is ‘‘medium risk aversion’’, and less than or equal to 3 is ‘‘high risk aversion.’’ As discussed earlier, people with lower wealth and base salary will generally be more averse to ﬁnancial risk since pay variability that results in pay reduction is more likely to force them to cut back on necessities, so we also examine the worker’s annual base salary and family wealth under the framework of risk aversion.

Risk Aversion Variables: Risk averse

Worker’s self-assessment of his or her risk preference, where 1 ¼ low risk aversion; 2 ¼ medium risk aversion; 3 ¼ high risk aversion

Base pay

Worker’s annual base pay the previous year excluding overtime, bonuses, and commissions (in thousands)

Wealth

Assets of the worker and the worker’s spouse including the value of their house minus the mortgage, their vehicles, stocks and mutual funds, cash, checking accounts, retirement accounts including 401(k), and pension assets (in thousands)

The second key element that is likely to shape attitudes toward ﬁnancial participation is residual control, or the ability to inﬂuence one’s potential for higher income, which will depend on worker skills and opportunities to inﬂuence workplace performance through employee involvement in decision making, training, and assurance that the worker can reap the rewards of higher performance through job security. These

Worker Attitudes Toward Employee Ownership

151

variables, which we group under the heading residual control, are deﬁned below. Residual Control Variables: Decision making

Dummy variable equaling 1 if the worker is involved in organized workplace decision making through teams, committees, or task forces that address workplace issues such as product quality, cost cutting, productivity, health, and safety; 0 otherwise

Training

Dummy variable equaling 1 if the worker received any formal training from the employer in the last 12 months, such as in classes or seminars sponsored by the employer; 0 otherwise

Job security Dummy variable equaling 1 if the worker’s response to the question ‘‘Thinking about the next 12 months, how likely do you think it is that you will lose your job or be laid off?’’ is ‘‘not at all likely’’ or ‘‘not too likely’’; 0 if the worker’s response is ‘‘very likely,’’ or ‘‘fairly likely’’.

As discussed earlier, preferences for company performance-based rewards in particular are likely to depend on workers’ perceptions about whether their co-workers are committed to workplace performance and whether they trust management to distribute payouts from workplace rewards correctly. We therefore include the below two independent variables. Perceptions of Co-Workers and Management: Co-worker interest Worker’s perception of his or her co-workers’ interest and involvement and involvement in company-wide issues on a 1–7 scale, with 1 indicating little interest and involvement and 7 indicating great interest and involvement Management Worker’s perception of the trustworthiness of his or trustworthiness her company in keeping its promises on a 0–4 scale, with 0 indicating not trustworthy and 4 indicating highly trustworthy

152

FIDAN ANA KURTULUS ET AL.

Control Variables Beyond the variables that are central to our analysis, all of our regression models include a rich array of worker and workplace characteristics as control variables. These are worker demographic variables including gender, age, and education level; and job characteristics variables including occupation, managerial level, ﬁrm tenure, whether the worker is closely supervised, and whether pay is at or above market level. We also control for the ease with which workers can observe their coworkers’ effort, since this is likely to inﬂuence the extent to which perceptions about how involved co-workers are in company issues (Coworker interest and involvement), one of the main independent variables we consider, affects preferences for shared capitalism. Fuller deﬁnitions and descriptive statistics of these control variables are provided in the appendix.

METHODOLOGY As a ﬁrst step in our empirical analysis, we will examine unconditional means of our main dependent variables to explore preferences over ﬁnancial participation broadly across all workers in our sample. Second, we will examine how preferences for ﬁnancial participation vary with worker risk aversion, residual control, and perceptions of co-workers and management by estimating cross-sectional regressions of attitudes on these main independent variables, controlling for the array of worker and workplace characteristics described earlier. We will estimate least squares, probit, and multinomial probit models, depending on whether the dependent variable is a continuous, dummy, or multivalued variable, respectively. It is important to note that these cross-sectional regression models do not capture causal relationships, but rather conditional correlations among the variables of interest. For example, it is possible that workers with a greater preference for ﬁnancial participation sort themselves into ﬁrms with greater use of ﬁnancial participation, greater residual control for workers, and workplace climates with high levels of trust. We investigated this possibility by also estimating regressions that included ﬁrm-ﬁxed effects, with very similar results for the variables of interest; here we report the results without ﬁrm effects to take advantage of both within- and

Worker Attitudes Toward Employee Ownership

153

between-ﬁrm variation. It also remains possible that within a ﬁrm, workers will be sorted into jobs based on personal characteristics correlated with preferences over ﬁnancial participation. It may be, for example, that being in a decision-making team does not create greater interest in ﬁnancial participation, but workers with greater interest in ﬁnancial participation select themselves into positions that are part of decision-making teams. Even in the latter case, ﬁnding a positive relationship strongly suggests an important linkage between residual control and residual returns for workers. So while we cannot deﬁnitively determine causality (as with most nonexperimental data), our results will nonetheless shed important light on how preferences for ﬁnancial participation are related to worker risk aversion, residual control and workplace climate, and the conditions under which variable pay plans are viewed positively by workers and are most likely to be effective.

UNCONDITIONAL STATISTICS Summary statistics for the main variables used in our empirical analyses are shown in Table 1, along with distribution charts for selected variables. Most workers desire between 0 and 30 percent of their compensation to be comprised of variable pay, though there is considerable variation in this preference across workers as illustrated in Fig. 1, with an average of 20 percent (Preference for variable pay). When asked whether workers prefer to be paid at least in part with variable pay that depends on company performance through proﬁt sharing, company stock or stock options, as opposed in getting all ﬁxed salary, a vast majority of respondents, 78 percent, say they prefer to have some company performance-dependent variable pay (Preference for company-based incentives). Only 27 percent of workers would like their next pay increase to come in the form of all ﬁxed wages with no proﬁt sharing, company stock or stock options, whereas 60 percent would like a combination of the two types, and 13 percent would like their next raise to consist entirely of proﬁt sharing, company stock or stock options (Preference for company-based incentives in next pay increase). Workers’ preferences in favor of company-performance-based pay is also evident in the distribution of the variable Preference for stock over cash incentives, where most workers picked categories 3 and 4 on a scale of 1–5 indicating their preference for getting some of their compensation from stock and stock options as opposed to a cash incentive plan.

154

FIDAN ANA KURTULUS ET AL.

Table 1.

Descriptive Statistics.

Panel A: Means, Standard Deviations, and Sample Sizes Variable

Dependent variables Preference for variable pay Preference for company-based incentives Preference for company-based incentives in next pay increase Preference for stock over cash incentives Risk aversion Risk averse Base pay Wealth Residual control Decision making Training Job security Perceptions of co-workers and management Co-worker interest and involvement Management trustworthiness

Mean

Standard Deviation

Observations

19.560 0.783 1.862

18.368 0.412 0.616

12,804 13,543 26,626

3.076

1.277

7,994

1.786 54.820 288.327

0.755 41.997 586.784

41,695 30,457 32,466

0.348 0.562 0.843

0.476 0.496 0.364

42,865 43,067 43,807

4.217 2.328

1.599 1.152

42,809 42,437

Panel B: Distribution Tables Frequency

Percent

1 (all ﬁxed wage) 2 3 (all company-performance-based pay)

7,143 16,014 3,469

26.83 60.14 13.03

Total

26,626

Preference for Company-Based Incentives in Next Pay Increase:

Preference for Stock Over Cash Incentives:

Frequency

1 (cash incentives) 2 3 4 5 (stock, options)

1,229 1,279 2,412 1,804 1,270

Total

7,994

Note: Based on the NBER Shared Capitalism Survey of N ¼ 46,907 workers.

100 Percent 15.37 16 30.17 22.57 15.89 100

155

0

.01

Density .02

.03

.04

Worker Attitudes Toward Employee Ownership

0

20

40

60

80

100

Preference for Variable Pay

Fig. 1. Kernel Density for Preference for Variable Pay. Note: Based on the NBER Shared Capitalism Survey of N ¼ 46,907 Workers. We Use an Epanechnikov Kernel with Bandwith 2.0128.

WHICH WORKERS PREFER FINANCIAL PARTICIPATION? The earlier discussion showed that most workers want at least a part of their compensation to be output-contingent, and prefer to be paid at least in part based on company performance. But it may well be that there is variation across these attitudes by the worker’s degree of risk aversion, residual control, and perceptions of co-workers and management. We now explore what types of workers prefer variable pay, and what kinds of variable pay they prefer, by estimating regressions of our various preference measures on the main independent variables of interest, and controls. Table 2 illustrates least squares regression results for Preference for variable pay, indicating the proportion of total pay the worker would like to receive as

156

FIDAN ANA KURTULUS ET AL.

Table 2. Relationship between Preference for Variable Pay and Risk Aversion, Residual Control, and Perceptions of Co-Workers and Management. Dependent Variable: Preference for Variable Pay Risk aversion Risk averse Base pay Wealth Residual control Decision making Training Job security Perceptions of co-workers and management Co-worker interest and involvement Management trustworthiness Worker and workplace controls Constant Observations Adjusted R2

4.173 (0.265) 0.051 (0.005) 0.002 (0.000) 0.974 (0.399) 0.671 (0.455) 0.626 (0.734) 0.427 (0.127) 0.720 (0.188) YES 10.580 (2.678) 8,289 0.220

Note: Results are from a least squares regression model. Robust standard errors in parentheses. Statistically signiﬁcant at the 0.10 level; at the 0.05 level; and at the 0.01 level. The model also includes the full set of worker and workplace controls described in the Data and Variables Section.

variable compensation, and provides strong support for our hypothesis on the relationship between risk aversion and preferences for variable pay: the proportion of compensation workers would like to receive as variable pay is negatively related with their degree of risk aversion – an increase in risk aversion from the ‘‘low’’ category to the ‘‘high’’ category is associated with a decrease in the desired proportion of pay comprised of variable compensation of over 8 percentage points, on average. Also, the proportion of compensation workers would like to receive as variable pay is statistically signiﬁcantly positively related with family wealth and base salary which can insulate

Worker Attitudes Toward Employee Ownership

157

workers against ﬂuctuations in compensation created by variable pay. Second, there is some support for the hypothesis that workers have greater preference for residual rewards when they also have residual control – the proportion of pay the worker desires in his or her compensation is signiﬁcantly positively related with employee involvement in workplace decision making, and is also positively related with formal job training and job security though these last two estimates are not statistically signiﬁcant at conventional levels. Third, in support of our hypothesis on perceptions of co-workers and management, preference for variable pay is positively related with co-worker interest and involvement in company-wide issues and trust in management, and both of these relationships are statistically signiﬁcant at the one-percent level, indicating that workers are more interested in shared capitalism when they can trust others in the workplace.1,2 Table 3 illustrates the probit marginal effects for Prob (Preference for company-based incentives ¼ 1), the probability that the worker prefers that he or she be paid in part with variable pay based on company performance (such as proﬁt sharing, company stock, and stock options) over being paid fully in the form of ﬁxed salary. We ﬁnd even stronger evidence in favor of our hypotheses here than in the previous table: preference for companyperformance-contingent pay is negatively associated with risk aversion, positively associated with family wealth and base salary, positively associated with all three measures of residual control, and positively associated with conﬁdence in co-workers and management, as predicted. We next turn to workers’ preferences over the portion of pay increases comprised of company-performance-based variable pay by focusing on the three-valued Preference for company-based incentives in next pay increase, which equals 1 if the worker prefers that his or her next pay increase is comprised of all ﬁxed wages, with no proﬁt sharing, company stock, or stock options; 2 if it is split between ﬁxed wages and proﬁt sharing, company stock, or stock options; and 3 if it is all in the form of proﬁt sharing, company stock, or stock options. Table 4 illustrates marginal effects from a multinomial probit regression of Preference for company-based incentives in next pay increase on our risk aversion, residual control, and co-worker and management perceptions variables, and worker controls, with the three columns corresponding to the probability that a worker chooses values 1, 2, or 3, respectively. We ﬁnd that workers who are more risk averse are less likely to want their next pay increase to be partly or completely comprised of company-performance-based pay, and more likely to prefer that it is entirely comprised of ﬁxed wages, corroborating our hypothesized relationship between risk aversion and attitudes toward ﬁnancial participation. On the

158

FIDAN ANA KURTULUS ET AL.

Table 3. Relationship between Preference for Company-Based Incentives and Risk Aversion, Residual Control, and Perceptions of Co-Workers and Management. Prob(Preference for Company-Based Incentives ¼ 1) Risk aversion Risk averse Base pay Wealth Residual control Decision making Training Job security Perceptions of co-workers and management Co-worker interest and involvement Management trustworthiness Worker and workplace controls Observations Pseudo R2

0.046 (0.005) 0.001 (0.000) 0.000 (0.000) 0.031 (0.007) 0.027 (0.008) 0.034 (0.013) 0.009 (0.002) 0.021 (0.003) YES 8,580 0.202

Note: Results are probit marginal effects for Prob(Preference for Company-Based Incentives ¼ 1) evaluated at the mean of the independent variable or, for binary independent variables, the change in the predicted Prob(Preference for Company-Based Incentives ¼ 1) when the independent variable increases from 0 to 1 (evaluating all other covariates at their means). Robust standard errors in parentheses. Statistically signiﬁcant at the 0.10 level; at the 0.05 level; and at the 0.01 level. The model also includes the full set of worker and workplace controls described in the Data and Variables Section.

contrary, wealth and salary do not exhibit statistically signiﬁcant relationships with Preference for company-based incentives in next pay increase. Table 4 provides mixed evidence supporting the notion that residual control improves preference for residual returns: job training and job security are associated with a higher likelihood of preferring one’s next pay increase to be comprised of both company performance related pay and ﬁxed wages, though the relationship between employee involvement in decision making

159

Worker Attitudes Toward Employee Ownership

Table 4. Relationship between Preference for Company-Based Incentives in Next Pay Increase and Risk Aversion, Residual Control, and Perceptions of Co-Workers and Management. Prob(Preference for Prob(Preference for Company-Based Company-Based Incentives in Next Pay Incentives in Next Pay Increase ¼ 1) Increase ¼ 2) (1) (2) Risk aversion Risk averse Base pay Wealth Residual control Decision making

0.052 (0.005) 0.000 (0.000) 0.000 (0.000)

0.012 (0.008) Training 0.021 (0.008) Job security 0.006 (0.012) Perceptions of co-workers and management Co-worker interest 0.001 and involvement (0.003) Management 0.026 trustworthiness (0.004) Worker and YES workplace controls Observations 13,363

Prob(Preference for Company-Based Incentives in Next Pay Increase ¼ 3) (3)

0.032 (0.006) 0.000 (0.000) 0.000 (0.000)

0.020 (0.004) 0.000 (0.000) 0.000 (0.000)

0.004 (0.009) 0.028 (0.009) 0.025 (0.013)

0.008 (0.005) 0.007 (0.005) 0.018 (0.008)

0.002

0.001

(0.003) 0.024

(0.002) 0.002

(0.004) YES

(0.003) YES

13,363

13,363

Note: Results in each column are multinomial probit marginal effects for Prob(Preference for Company-Based Incentives in Next Pay Increase ¼ 1), Prob(Preference for Company-Based Incentives in Next Pay Increase ¼ 2), and Prob(Preference for Company-Based Incentives in Next Pay Increase ¼ 3), respectively, evaluated at the mean of the independent variable or, for binary independent variables, the change in the predicted probability when the independent variable increases from 0 to 1 (evaluating all other covariates at their means). Robust standard errors in parentheses. Statistically signiﬁcant at the 0.10 level; at the 0.05 level; and at the 0.01 level. The models also include the full set of worker and workplace controls described in the Data and Variables Section.

160

FIDAN ANA KURTULUS ET AL.

and Preference for company-based incentives in next pay increase is not statistically signiﬁcant. Finally, though trust in management is associated with wanting a portion of one’s pay raise to be contingent on company performance, co-worker interest and involvement in company-wide issues does not have a statistically signiﬁcant relation with Preference for companybased incentives in next pay increase. Table 5 shows results from a least squares regression of preference for receiving some compensation from company stock and stock options as

Table 5. Relationship between Preference for Stock Over Cash Incentives and Risk Aversion, Residual Control, and Perceptions of Co-Workers and Management. Dependent Variable: Preference for Stock Over Cash Incentives Risk aversion Risk averse Base pay Wealth Residual control Decision making Training Job security Perceptions of co-workers and management Co-worker interest and involvement Management trustworthiness Worker and workplace controls Constant Observations Adjusted R2

0.231 (0.022) 0.002 (0.000) 0.000 (0.000) 0.140 (0.031) 0.102 (0.037) 0.175 (0.058) 0.075 (0.010) 0.162 (0.015) YES 1.373 (0.248) 6,766 0.151

Note: Results are from a least squares regression model. Robust standard errors in parentheses. Statistically signiﬁcant at the 0.10 level; at the 0.05 level; and at the 0.01 level. The model also includes the full set of worker and workplace controls described in the Data and Variables Section.

Worker Attitudes Toward Employee Ownership

161

opposed to a cash incentive plan on a 1 to 5 scale (Preference for stock over cash incentives). We can interpret this as capturing whether the respondent prefers having stock or cash in his or her pocket. The worker’s selfassessment of his or her own risk aversion is strongly negatively correlated with preference for stock-based compensation as opposed to cash incentives, as expected. However, family wealth and base salary have a statistically signiﬁcant positive, albeit small, correlation with preference for stock-based compensation over cash incentives. All of our residual control variables and management and co-worker trust variables are positively related to Preference for stock over cash incentives, in support of our hypotheses.3

CONCLUSION This study uses the NBER Shared Capitalism Survey to examine a relatively unexplored topic in the research on participatory pay schemes, namely worker preferences for employee ownership, proﬁt sharing and variable pay, and how these preferences depend on three key worker and workplace characteristics: worker risk aversion, residual control, and perceptions of coworkers and management. We ﬁnd that, on average, workers desire around 20 percent of their compensation to be comprised of variable pay. Most workers prefer to be paid at least in part based on company performance, through proﬁt sharing, company stock, or stock options, and in particular they prefer getting stock and stock options as opposed to cash incentives. Furthermore, our regression results clearly indicate that risk aversion is a major factor reducing preferences for variable pay plans. This is supported not only by the strong coefﬁcient estimates on the risk aversion variable, but also by the coefﬁcient estimates on two key employee characteristics that are expected to be related to risk preferences, namely base pay and family wealth. An important ﬁnding, though, is that workplace policies help to improve worker perceptions of variable pay. The ﬁnding that workers are more likely to prefer performance-based pay if they have decision-making power, formal job training, and job security at their workplace supports the theory that residual control and residual returns are complementary. Future research would be valuable on the form of this relationship (e.g., linear or nonlinear, including whether there are threshold effects), and whether complementary residual control practices can help make shared capitalism more appealing to groups who are more risk-averse.

162

FIDAN ANA KURTULUS ET AL.

Workplace culture also is a key variable. Workers are unlikely to favor variable pay plans if they do not trust managers (either to manage well so that there will be rewards, or to calculate rewards accurately and honestly). Their attitudes are also inﬂuenced by how they perceive their co-workers: if they have little reason to expect their co-workers to perform well under a performance-based pay plan, they will not be optimistic about the prospect for rewards. The ﬁndings on workplace policies and trust of co-worker and managers indicate that preferences over variable pay plans are not determined by any ﬁxed mindset or personal characteristic of workers, but appear to depend on the context in which the plans are implemented. This is consistent with research on the performance effects of employee ownership and proﬁt sharing, which shows that while these plans are associated with higher performance effects on average, there is substantial dispersion in estimated effects across and within samples (Doucouliagos, 1995; Freeman, 2007; Kaarsemaker, 2006). This dispersion is likely to be explained in part by employees’ neutral or negative reactions in workplaces that do not provide supportive environments, in contrast to positive reactions with increased employee effort and cooperation when the environment is supportive. Research has begun to identify how speciﬁc workplace policies condition the effects of group incentive plans (Kruse, Blasi, & Park, 2010). Further research that delves into the role of workplace policies and cultures in ﬁnancial participation would be valuable in determining how and when these plans can affect performance and worker welfare, and the likelihood that they may expand in the 21st century.

NOTES 1. Apart from these main results on how preferences for variable pay depend on risk aversion, residual control and perceptions of co-workers and management, we also uncovered some interesting ﬁndings on the relationships between the control variables and Preference for variable pay. For example, women have a lower preference for variable pay, which aligns with evidence from past studies using laboratory and ﬁeld experiments that women tend to be more risk averse than men (Dohmen & Falk 2011; Dohmen et al., 2011; Niederle & Vesterlund, 2007). We also ﬁnd that workers in sales and customer service occupations have the strongest preference in favor of variable pay, a possible explanation for which is that output is more readily linkable to individual performance than for workers in many other occupations (e.g., number of units sold or number of customers assisted), and this reduces problems of free-riding so that such workers may view output-contingent compensation as a more fruitful reward for their effort. Among other interesting

Worker Attitudes Toward Employee Ownership

163

results is that workers with longer tenure at the ﬁrm prefer a lower share of variable pay, supporting the idea that workers who are closer to retirement are often loathe to introduce risk into their compensation as they have less time remaining in the labor market to recoup potential losses. To streamline and focus our discussion around the key variables of interest (risk aversion, residual control, and perceptions of coworkers and management) we do not present these results here, but they are available from the authors. 2. Since Preference for variable pay has a lower bound of zero we also estimated an analogous regression using a tobit model and obtained very similar results. 3. As a robustness check we also estimated all of our regressions with ﬁrm-ﬁxed effects, and the results were very similar to the baseline results we report here, both in magnitude and signiﬁcance. These additional results are available from the authors.

REFERENCES Axelrod, R. (1984). The evolution of cooperation. New York: Basic Books. Ben-Ner, A., & Jones, D. C. (1995). Employee participation, ownership, and productivity: A theoretical framework. Industrial Relations, 34(4), 532–554. Ben-Ner, A., Kong, F., & Lluis, S. (2010). Uncertainty and organization design. Working paper, University of Minnesota. Brown, S., & Sessions, J. G. (2003). Attitudes, expectations and sharing. Labour, 17(4), 543–569. Cadsby, C. B., Song, F., & Tapon, F. (2007). Sorting and incentive effects of pay for performance: An experimental investigation. Academy of Management Journal, 50(2), 387–405. Cornelissen, T., Heywood, J. S., & Jirjahn, U. (2008). Performance pay, risk attitudes and job satisfaction. DIW Working Paper. Craig, B., & Pencavel, J. (1992). The behavior of worker cooperatives: The plywood companies of the Paciﬁc Northwest. American Economic Review, 82, 1083–1105. Del Boca, A., Kruse, D., & Pendelton, A. (1999). Decentralisation of bargaining systems and ﬁnancial participation: A comparative analysis of Italy, U.K. and the U.S. Lavoro e Relazioni Industriali, Summer. DeVaro, J., & Kurtulus, F. A. (2010). An empirical analysis of risk, incentives, and the delegation of worker authority. Industrial and Labor Relations Review, 63(4), 641–661. Dohmen, T., & Falk, A. (2011). Performance pay and multidimensional sorting: Productivity, preferences and gender. American Economic Review, 101(2), 556–590. Dohmen, T., Falk, A., Huffmann, D., Sunde, U., Schupp, J., & Wagner, G. (2011). Individual risk attitudes: Measurement, determinants and behavioral consequences. Journal of the European Economic Association, 9(3), 522–550. Doucouliagos, C. (1995). Worker participation and productivity in labor-managed and participatory capitalist ﬁrms: A meta-analysis. Industrial and Labor Relations Review, 49(1), 58–77. Drago, R., Estrin, S., & Wooden, M. (1992). Pay for performance incentives and work attitudes. Australian Journal of Management, 17(2), 217–231. Foss, N. J., & Laursen, K. (2005). Performance pay, delegation and multitasking under uncertainty and innovativeness: An empirical investigation. Journal of Economic Behavior and Organization, 58, 246–276.

164

FIDAN ANA KURTULUS ET AL.

Freeman, S. F. (2007). Effects of ESOP adoption and employee ownership: Thirty years of research and experience. Working Paper no. 07-01, Organizational Dynamics Programs, University of Pennsylvania. Fudenberg, D., & Maskin, E. (1986). The folk theorem in repeated games with discounting or with incomplete information. Econometrica, 54(3), 533–554. Green, C., & Heywood, J. (2008). Does performance pay increase job satisfaction? Economica, 75, 710–728. Holmstrom, B. (1979). Moral hazard and observability. Bell Journal of Economics, 10(1), 74–91. Holmstrom, B., & Milgrom, P. (1994). The ﬁrm as an incentive system. American Economic Review, 84(4), 972–991. Jensen, M. C., & Meckling, W. H. (1992). Speciﬁc and general knowledge, and organizational structure. In: L. Werin & H. Wijkander (Eds.), Contract economics. Oxford, UK: Blackwell Publishers. Jones, D. C., & Kato, T. (1995). The productivity effects of employee stock-ownership plans and bonuses: Evidence from Japanese panel data. American Economic Review, 85(3), 391–414. Kaarsemaker, E. C. A. (2006). Employee ownership and its consequences: Synthesis-generated evidence for the effects of employee ownership and gaps in the research literature. York, UK: University of York. Kruse, D. (2002). Research evidence on prevalence and effects of employee ownership. Testimony before the Subcommittee on Employer-Employee Relations, Committee on Education and the Workforce, U.S. House of Representatives, February 13. Kruse, D., & Blasi, J. R. (1997). Employee ownership, employee attitudes, and ﬁrm performance: A review of the evidence. In: D. Lewin, D. J. B. Mitchell & M. A. Zaidi (Eds.), The human resource management handbook, part I. Greenwich, CT: JAI Press Inc. Kruse, D., & Blasi, J. R. (1999). Public opinion polls on employee ownership and proﬁt sharing. Journal of Employee Ownership Law and Finance, 11(3), 3–25. Kruse, D. L., Blasi, J. R., & Park, R. (2010). Shared capitalism in the U.S. economy: Prevalence, characteristics, and employee views of ﬁnancial participation in enterprises. In: D. Kruse, R. B. Freeman & J. R. Blasi (Eds.), Shared capitalism at work: Employee ownership, oﬁt and gain sharing, and broad-based stock options. Chicago: The University of Chicago Press. Levine, D., & Tyson, L. (1990). Participation, productivity and the ﬁrm’s environment. In: A. Blinder (Ed.), Paying for productivity: A look at the evidence. Washington, DC: Brookings Institution. McCausland, W. D., Pouliakas, K., & Theodossiou, I. (2005). Some are punished and some are rewarded: A study of the impact of performance pay on job satisfaction. International Journal of Manpower, 26(7/8), 636–659. Milgrom, P., & Roberts, J. (1990). Economics, organization and management. New York: Prentice Hall. Niederle, M., & Vesterlund, L. (2007). Do women shy away from competition? Do men compete too much? Quarterly Journal of Economics, 122(3), 1067–1101. Park, R., Kruse, D., & Sesil, J. (2004). Does employee ownership enhance ﬁrm survival? In: V. Perotin & A. Robinson (Eds.), Advances in the economic analysis of participatory and labor-managed ﬁrms (Vol. 8, pp. 3–33). New York: Elsevier Science, JAI. Prendergast, C. (2002). The tenuous tradeoff between risk and incentives. Journal of Political Economy, 110(5), 1071–1102.

Worker Attitudes Toward Employee Ownership

165

Shavell, S. (1979). Risk sharing and incentives in the principal and agent relationship. Bell Journal of Economics, 10(1), 55–73. Weitzman, M., & Kruse, D. (1990). Proﬁt sharing and productivity. In: A. Blinder (Ed.), Paying for productivity: A look at the evidence. Washington, DC: Brookings Institution. Zalusky, J. L. (1986). Labor’s collective bargaining experience with gainsharing and proﬁt sharing. IRRA 39th Annual Proceedings, 174–182. Zalusky, J. L. (1990). Labor-management relations: Unions view proﬁt sharing. In: M. J. Roomkin (Ed.), Proﬁt sharing and gain sharing (pp. 65–78). Metuchen, NJ: Scarecrow Press.

166

FIDAN ANA KURTULUS ET AL.

APPENDIX

Deﬁnitions and Descriptive Statistics of Worker and Workplace Control Variables. Variable

Deﬁnition

Mean

Standard Deviation

Observations

Female

1 if worker is female; 0 otherwise Worker age 1 if worker does not hold a high school degree; 0 otherwise 1 if worker’s highest educational degree is a high school degree including GED; 0 otherwise 1 if worker has attended some college but has not received a bachelor’s degree; 0 otherwise 1 if worker’s highest educational degree is an associate’s degree; 0 otherwise 1 if worker’s highest educational degree is a bachelor’s degree; 0 otherwise 1 if worker’s highest educational degree is a master’s, professional or doctoral degree; 0 otherwise 1 if worker’s occupation is production; 0 otherwise

0.312

0.463

38,325

40.933 0.037

10.503 0.189

36,791 35,758

0.230

0.421

35,758

0.217

0.412

35,758

0.084

0.277

35,758

0.280

0.449

35,758

0.138

0.344

39,436

0.434

0.496

45,816

Age No high school High school

Some college

Associate degree

College

Grad school

Production

167

Worker Attitudes Toward Employee Ownership Administrative support Professional and technical

Sales and customer service Lower management

Middle management

Upper management

At market salary

Tenure Hours

1 if worker’s occupation is administrative support; 0 otherwise 1 if worker’s occupation is professional and technical (including engineers and scientists); 0 otherwise if worker’s occupation is sales and customer service; 0 otherwise 1 if worker’s occupation is lower management (including front-line supervisors); 0 otherwise 1 if worker’s occupation is middle management (including managers and directors); 0 otherwise 1 if worker’s occupation is top management (executives); 0 otherwise 1 if the worker believes that his or her annual annual base salary at the ﬁrm is at or above the going market rate for employees in other companies with similar experience and job descriptions in the region; 0 otherwise Worker’s tenure at the ﬁrm, in years Worker’s weekly hours worked

0.061

0.238

45,816

0.295

0.456

45,816

0.085

0.280

45,816

0.101

0.302

45,816

0.075

0.263

45,816

0.022

0.147

45,816

0.592

0.491

36,236

9.540

8.979

45,755

45.789

8.137

45,696

168 Close supervision

See co-workers

FIDAN ANA KURTULUS ET AL. Measure of how closely the worker is supervised on a 0–10 scale, with 0 indicating that the worker works independently of close supervision and 10 indicating the worker is closely supervised Worker’s rating of how easy it is for him to see whether his co-workers are working well or poorly on a 1–10 scale

3.347

2.631

45,978

6.784

2.740

45,874

Note: Based on the NBER Shared Capitalism Survey of N ¼ 46,907 workers.

PERFORMANCE-RELATED PAY, UNIONS, AND PRODUCTIVITY IN ITALY: EVIDENCE FROM QUANTILE REGRESSIONS Mirella Damiani and Andrea Ricci ABSTRACT This chapter explores the relationship between performance-related pay (PRP) and productivity in the Italian economy. It contributes to the literature in two main ways. First, it provides estimates for the PRP – productivity relationship based on a nationally representative sample of manufacturing and services companies (other studies on Italy are more limited in scope, since they focus on speciﬁc sectors). Second, it addresses the question of ﬁrms’ heterogeneity, an aspect so far not examined in relation to PRP in Italy. We use a two-step approach. In the ﬁrst step, we estimate a classical production function using longitudinal data on the balance sheet variables of Italian ﬁrms over the period 2002–2005. In the second step, we regress the distribution of the ﬁrm-speciﬁc ﬁxed effects on dummy variables for the presence of PRP and unions, as well as on control variables for the year 2005. The most important results are that the adoption of PRP is positively and uniformly correlated with productivity throughout the whole distribution and that the presence of trade unions has a positive association with ﬁrms’ unobserved productivity across all

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 169–196 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012011

169

170

MIRELLA DAMIANI AND ANDREA RICCI

quantiles, being signiﬁcantly higher for the best performing ﬁrms (those placed at the highest quantile of the productivity distribution). Keywords: Performance-related pay; productivity; unions JEL classiﬁcations: J33; D24; J51

INTRODUCTION This chapter explores the relationship between performance-related pay (PRP) and productivity in the Italian economy, using a unique dataset that collects information at ﬁrm level for both manufacturing and services sectors. It contributes to the literature in two main ways. First, it provides estimates for the PRP – productivity relationship based on a nationally representative sample of manufacturing and services companies. Until now, all available research has been restricted to large companies, selected sectors or particular areas in the north of the country (see Origo, 2009 for a summary), while no studies have considered this issue on a national scale. Second, it addresses the question of ﬁrms’ heterogeneity, so far not examined in relation to PRP in Italy and considered in only a limited number of other European countries. In fact, only a few empirical studies on wage rules, productivity and heterogeneity in Europe have been carried out (e.g., Bastos, Monteiro, & Straume, 2009, for Portugal; Wagner, Addison, Claus Schnabel, & Schank, 2004, for Germany). We adopt a two-step estimation procedure, similar to that used by Black and Lynch (2001), together with quantile regressions. In the ﬁrst step, we estimate a classical production function using longitudinal data on the balance sheet variables of Italian ﬁrms over the period 2002–2005. The distribution of average ﬁrm residuals (obtained by the ﬁrst step) is then used as a measure of the ﬁrm ﬁxed effect and regressed, in the second step, using quantile methods on our 2005 data for PRP, unionization and other ﬁrm and workplace characteristics. Two main results emerge: ﬁrst, the adoption of PRP is positively and uniformly associated with productivity throughout the whole distribution; second, the presence of unions has a positive association with ﬁrms’ unobserved productivity across all quantiles, being signiﬁcantly higher for the best performing ﬁrms (those placed at the highest quantile of the productivity distribution).

Performance-Related Pay, Unions, and Productivity in Italy

171

Italy is an interesting case study, since a process of large-scale reform of its industrial relations system began in the early 1990s. However, since the mid-1990s, Italian labor productivity growth has recorded a signiﬁcant slowdown. Also, as reported by Daveri and Jona-Lasinio (2005), growth accounting has revealed the crucial inﬂuence of total factor productivity (TFP), the component of productivity that depends not on capital deepening but on organizational strategies. Over the same period, similar labor market reforms have been introduced in many other European countries, with the aim of favoring ‘‘ﬂexibility’’ in industrial relations and linking wage increases to the dynamics of labor productivity. This type of reform raises many issues (OECD, 2004, Chapter 3). One concern is related to the expected positive effects of wage ﬂexibility and PRP on productivity, as well as to the possible negative side-effects caused by rent-sharing, facilitated by agreements at ﬁrm level. This theme has stimulated new interest in recent empirical literature. In two European economies, Germany (Gu¨rtzgen, 2009) and Belgium (Rusinek & Rycx, 2008), it has been ascertained that the disadvantages of distributive, rentsharing rules outweigh the positive effects of ﬂexible wage structures. Another critical issue regards the strategic role of complementary workplace practices which may inﬂuence the effects of PRP. For instance, Black and Lynch (2001) ﬁnd that unionized plants in the American economy, where joint decision making accompanies incentive payments, record higher productivity performance than non-unionized plants. The adoption of ‘high involvement’ systems would thus seem to be a necessary condition for making contingent pay settings effective on productivity growth, as shown by a growing body of empirical literature (see for a survey Godard & Delaney, 2000). Micro-evidence reveals that disparities in efﬁciency gains may persist even among ﬁrms within a single country. Indeed, it is now widely recognized (after the contribution on international trade by Melitz, 2003) that ﬁrms are heterogeneous with respect to key variables, including productivity and wage setting systems. The controversial inﬂuence of PRP thus seems to ﬁnd its natural place within the literature on heterogeneity. The present study integrates these elements – the role of PRP and ﬁrms’ heterogeneity – and tries to ﬁll this gap. This paper is organized as follows. The second section brieﬂy discusses the theoretical and empirical literature. The third section presents the Italian data that has been used and offers descriptive statistics. The fourth section illustrates the econometric framework. The ﬁfth section presents the estimation results, and the sixth section concludes.

172

MIRELLA DAMIANI AND ANDREA RICCI

PRODUCTIVITY, PERFORMANCE PAYMENTS, AND UNIONS: A SHORT REAPPRAISAL Our research is related to various ﬁelds of theoretical and empirical literature. One is that of proﬁt-sharing, more recently known as ‘‘employee ﬁnancial participation,’’ which posits that payments of collective bonuses, such as proﬁt-sharing, constitute a commitment device to motivate groups of workers and their collaborative relationships.1 Proﬁt sharing is held to generate beneﬁcial effects in the form of higher effort and work quality, higher commitment and incentives to ﬁrm-speciﬁc human capital, better teamwork, greater workforce cooperation in facing new technology and organizational changes, lower labor turnover, and longer average tenure (see, among others, the contributions of Estrin, Grout & Wadhwani, 1987; Jones & Pliskin, 1991; Svejnar, 1982). It is argued that these collective incentive schemes are more likely to be offered when total output is the result of the efforts of many agents and individual contributions cannot easily be identiﬁed (Fitzroy & Kraft, 1992; Holmstrom, 1979). In such cases, the absence of group incentives may lead to inferior Nash equilibria, associated with low levels of productivity due to limited cooperation. Conversely, offering collective incentives in the form of contractual shares of companies motivates workers to high levels of cooperation and allows lower monitoring costs (Fitzroy & Kraft, 1987). Collective bonuses are nevertheless not exempt from potential drawbacks. The very fact that they are collective may induce employees to free-ride on the efforts of others and thus cut productivity. Horizontal mutual monitoring and peer pressure among workers may be a solution to this problem (Kandel & Lazear, 1992). Freeman, Kruse, and Blasi (2010) have recently put forward evidence that employees believe they are able to monitor the activities of their fellows, a precondition of peer pressure. Moreover, employees in ﬁrms with ‘shared capitalism’ (i.e., those with proﬁtsharing and/or employee share ownership) respond more readily to the shirking of their colleagues Similar empirical conﬁrmation is given by Mas and Moretti (2009), who ﬁnd that workers whose efforts may be noticed by their fellows display more cooperative attitudes. Negative externalities (which are pervasive in workplaces), are thus internalized not because of altruistic behavior, but because peer pressure discourages free-riding, especially when workers expect that many future interactions with the same peers will occur. More generally, a plausible solution to the quelling of free-riding attitudes is the promotion of team culture and employee participation in decision

Performance-Related Pay, Unions, and Productivity in Italy

173

making, a policy that contributes, like ﬁnancial participation, to increasing commitment (Blinder, 1990; Jones & Pliskin, 1991;2 Kruse, Freeman, & Blasi, 2010). On empirical grounds, a growing body of research suggests that the PRP– productivity relationship may be positive, as shown by the ﬁrst report of the experience in European countries (Uvalic, 1991), as well as other studies for major industrialized economies (see, among others, Kruse, 1993; Pe´rotin & Robinson, 2003). Experiences in more than 20 countries offer empirical evidence of the positive or, at least, neutral effects on productivity, and the discrepancies may be attributed ‘‘to differences in participatory practices in ﬁrms with proﬁt-sharing plans’’ (Pe´rotin & Robinson, 2003, p. 22).3 Another ﬁeld of literature surmises that individual bonuses linked to performance may be an efﬁcacious device to motivate single employees. These studies, appraised by Prendergast (1999), point out that the ‘‘power of incentives’’ depends above all on motivation and sorting. Pay settings that change from rewards based on input measures to payments related to output outcomes may induce dramatic improvements, half of which are explained by the attraction of workers of higher ability (Lazear, 2000). The same ﬁeld of literature has paid close attention to interactions with other packages of good workplace practices, giving rise to many studies exploring the ‘‘value of the complementary role of human resource practices’’ on nationally representative samples of enterprises (Black & Lynch, 2001) or insider econometric case studies within ﬁrms (Ichniowski & Shaw, 2003). Controversial aspects of incentives are highlighted by Bandiera, Barankay, and Rasul (2005), who show that employees underperform when wages are linked to relative performance, since workers whose efforts impose negative externalities on their fellows internalize fear retaliation. The consequence is that the average worker’s productivity is signiﬁcantly lower under relative incentives than under piece rates. In another study, Bandiera, Barankay, and Rasul (2010) examine the importance of social ties across workers and ﬁnd positive spillover effects where social ties exist, as a given worker’s productivity is signiﬁcantly higher when that person works together with friends, especially those who are more able. To motivate workers, ﬁrms may therefore choose to exploit social incentives as an alternative to monetary incentives. Another controversial aspect arises from the possible trade-off between extrinsic and intrinsic motivations, since contingent rewards may conﬂict with intrinsic motivation, so impairing performance (Benabou & Tirole, 2003).

174

MIRELLA DAMIANI AND ANDREA RICCI

A third body of literature considers a problematic line of inquiry, namely the collective actions of workers and their representatives. It has been argued that, in conditions of imperfect competition and high levels of unionization, company wage agreements cause the redistribution of rents, as well as collusive behavior between management and employees. Empirical evidence is offered by Hu¨bler and Jirjahn (2003), who ﬁnd that, in German ﬁrms, in the case of company negotiations with unions, rent-sharing effects are more pronounced. Another view emphasizes the positive inﬂuence of unions, whose presence tends to generate improvements in productivity, decreases in the dispersion of earnings and reserves greater space for the workers’ voice (see the ample discussion in Addison & Hirsch, 1989). Nevertheless, the effects of unions on pay settings and productivity are ambiguous, and it is not surprising that international evidence shows contradictory ﬁndings. For instance, in the United States, Black and Lynch (2001) estimate that unionized establishments that adopt incentive-based compensations (associated with joint decision making) have higher productivity than other similar non-union plants. In our case study, Italy, Origo (2009) reports opposite results for the metal-working sector. She estimates the effects of a shift in a ﬁrm’s pay strategy from a ﬁxed to a ﬂexible wage, based on panel data, for a representative sample of Italian metal-working ﬁrms in the 1990s. She ﬁnds that, in line with theoretical predictions, the ‘‘treated’’ ﬁrms, that is, ﬁrms adopting PRP, have positive effects on labor productivity growth (of around 7–11%), with beneﬁcial effects shared with employees in the form of higher wages (around 2–3%). However, her estimates suggest that productivity effects are greater in low-unionized ﬁrms, whereas the opposite is true for remuneration, thus showing that high-unionized companies are more oriented toward rent-sharing. The dataset used in this chapter, which gives information on the workplace practices in a single year (2005), does not enable us to identify causal effects, like those obtained with the Propensity Score Matching estimator used by Origo. However, our unique dataset contains more recent information and a nationwide sample of ﬁrms, representative of the whole Italian economy, unlike Origo’s data, which refers to the 1990s and to one sector only. The data we use cover all non-agricultural sectors and companies of all sizes, as well as containing a wealth of information on ﬁrm and employee characteristics (see the third section). The same data were used in a previous study (Damiani & Ricci, 2009) aimed at identifying the main factors leading to the adoption of PRP. The study found marked differences across sectors and regions and a greater

Performance-Related Pay, Unions, and Productivity in Italy

175

presence of unions in PRP ﬁrms, as opposed to non-PRP ﬁrms. This paper integrates our 2009 analysis, addressing the efﬁciency issue and estimating the extent to which the PRP–productivity link is affected by ﬁrm timeinvariant heterogeneities.

DATA AND DESCRIPTIVE STATISTICS Our empirical study is based on a nationally representative sample of manufacturing and non-manufacturing ﬁrms, obtained by merging information from two different sources: balance-sheet data from the Bureau Van Dijk AIDA archive and ﬁrm-level information on PRP and other workplace practices from the ISFOL Employer and Employee Survey (RIL). The ISFOL-RIL survey is a ﬁrm survey that collected cross-sectional information relative to 2005 about personnel organization, recruitment strategies, position of employees, training investments, presence of unions, adoption of PRP schemes and other workplace characteristics. It refers to ﬁrms operating in the non-agricultural private sector and includes both partnership and limited companies, for a total sample of 21,728 ﬁrms. As far as PRP is concerned, each ﬁrm is asked whether or not such a scheme is adopted. Unfortunately, we do not know whether the different types of schemes are based on ﬁrm-, group-, or individual-performance. Besides, the dataset does not provide statistics on how many workers in the ﬁrm receive PRP or whether these schemes are offered to all or to a selected group of employees (managers, blue-collars, or all workers). Therefore, our PRP-variable is a dummy variable simply indicating the existence or not of a PRP scheme of some kind. As regards unions, the respondent ﬁrm is asked whether there is a form of employee representation of any kind in the ﬁrm. We thus have a second dummy variable indicating the presence of unions at ﬁrm level. Furthermore, we have information about the workplace characteristics and business strategies of each ﬁrm (detailed deﬁnitions of all variables are given in the Table A1). The AIDA database contains the annual accounts for limited companies that had turnovers of over 100,000 Euros in 2004 (the turnover threshold was previously 500,000 Euros). This database is a source of information on value added, capital, labor, and R&D for the period between 1997 and 2005 (see the Appendix). To link information concerning workers’ characteristics to indicators of ﬁrm performance and accounting variables, a sub-sample of the RIL dataset

176

MIRELLA DAMIANI AND ANDREA RICCI

was merged, as said above, with balance-sheet information from the AIDA archive for a period of four years (2002–2005), using company tax codes. Thus, the merged RIL-AIDA sample contains cross-sectional information about employees’ participation and other workplace practices for 2005, and the longitudinal accounting data for the period 2002–2005. Given the characteristics of the RIL-AIDA dataset, the merged sample is representative only in the case of limited companies. Also, we exclude ﬁrms with fewer than ﬁve employees, applying a ﬁlter to retain only those characterized by a minimum level of organizational structure. After matching and data validation, we obtained an unbalanced panel of 6,160 ﬁrms. Descriptive Statistics In this section, we present descriptive statistics of the AIDA-RIL merged sample for the year 2005. In view of our focus on heterogeneity, we single out three groups of ﬁrms, according to their average productivity performance over the period 2002–2005: ‘‘low performers’’ fall in the group between the 1st and 25th percentiles of the productivity distribution, ‘‘middle performers’’ in the group of ﬁrms whose productivity is between the 50th and 75th percentiles of the average distribution, and ‘‘high performers’’ are those ﬁrms with productivity higher than the 75th percentile. We then examine the main characteristics of these different groups of ﬁrms and of the whole sample. As regards enterprise characteristics, we examine the value added, ﬁxed capital, a dummy variable indicating whether the ﬁrms compete in the international market. We group ﬁrms in four classes (5–9, 10–49, 50–249, and more than 250 employees), four geographical macro-areas (the NorthWest, North-East, Central, and Southern regions of Italy) and eight 2-digit sectors.4 In the case of workers characteristics, we consider the percentage of women, the professional composition of employees (managers and supervisors, white- and blue-collar workers), the proportion of ﬁxed-term contracts and trained workers. For workplace characteristics we include two dummy variables indicating the adoption of PRP and the presence of unions at ﬁrm level, respectively. We also take into account the local unemployment rate and the vacancy rate for each ﬁrm, principally to control labor market tightness. Table 1 reports the summary statistics for these variables. First, we consider a group of variables related to three ﬁrm strategies: capital accumulation, internationalization, and incentive wage settings. The

177

Performance-Related Pay, Unions, and Productivity in Italy

Table 1.

Statistics by Firm Performance – 2005.

Variable

Low

Medium

High

Whole Sample

Mean SD Mean SD Mean SD Mean SD PRP Average (log) value added 2002–2005 Average (log) ﬁxed capital 2002–2005 Union % Women % Managers % White collars % Blue collars % Trained % Fixed-term contracts Foreign Unemployment rate 2004 Vacancy rate Firm size No. of employees o10 No. of employees W9 and o50 No. of employees W49 and o250 No. of employees W249 and more Macro-region North-West North-East Centre South Sector Mining, distribution of gas, and others Manufacturing Construction Trade, hotels, and restaurants Transport and communications Financial intermediation Other business services Education, health, and other business service No. of observations

0.01 7.82 6.63 0.03 0.41 0.04 0.45 0.47 0.12 0.13 0.34 7.81 0.15

0.10 0.56 1.39 0.17 0.30 0.09 0.34 0.34 0.26 0.20 0.47 4.93 0.38

0.08 9.25 8.31 0.24 0.33 0.03 0.46 0.46 0.19 0.09 0.46 6.49 0.06

0.27 0.43 0.50 0.04 0.19 0.50 11.27 0.73 8.60 1.14 1.46 10.33 1.51 7.52 1.80 0.43 0.72 0.45 0.09 0.29 0.25 0.31 0.22 0.46 0.34 0.06 0.03 0.06 0.04 0.11 0.30 0.48 0.29 0.45 0.35 0.30 0.44 0.29 0.46 0.35 0.29 0.22 0.29 0.12 0.27 0.13 0.09 0.12 0.14 0.22 0.50 0.67 0.47 0.31 0.46 4.20 5.71 3.30 7.29 4.72 0.11 0.01 0.04 0.16 0.36

0.58 0.42 0.00 0.00

0.49 0.49 0.05 0.01

0.10 0.78 0.12 0.01

0.29 0.41 0.32 0.08

0.00 0.18 0.62 0.20

0.05 0.38 0.49 0.40

0.58 0.36 0.05 0.01

0.49 0.48 0.21 0.09

0.35 0.10 0.29 0.27

0.48 0.30 0.45 0.44

0.46 0.14 0.25 0.16

0.50 0.34 0.43 0.36

0.56 0.14 0.20 0.10

0.50 0.35 0.40 0.30

0.40 0.11 0.27 0.22

0.49 0.31 0.44 0.42

0.00 0.32 0.19 0.33 0.05 0.00 0.07 0.03

0.06 0.47 0.39 0.47 0.21 0.06 0.26 0.16

0.01 0.46 0.14 0.20 0.06 0.00 0.08 0.04

0.11 0.50 0.35 0.40 0.24 0.07 0.28 0.20

0.01 0.55 0.04 0.16 0.06 0.00 0.12 0.05

0.12 0.50 0.20 0.37 0.24 0.04 0.32 0.21

0.01 0.27 0.15 0.28 0.04 0.01 0.19 0.04

0.08 0.45 0.36 0.45 0.20 0.11 0.39 0.20

1,380

3,180

1,592

6,160

Notes: ‘‘Low’’ is the group of ﬁrms whose performances over the period 2002–2005 are minor of the 25th percentile of the average value added per employee distribution, ‘‘Medium’’ indicates the group with past performances between the 25th and 75th percentiles, and ‘‘High,’’ the group with past performances over the 75th percentile.

178

MIRELLA DAMIANI AND ANDREA RICCI

typical proﬁle of high-performer ﬁrms emerges clearly from our data: they are more active in terms of capital accumulation, more frequently represented in international markets (as recent literature on internalization has shown, Mayer & Ottaviano, 2007), and more oriented toward adopting incentive pay systems. The opposite is true for low performers, while companies in the intermediate group occupy an intermediate position in terms of capital accumulation, exposure to international competition, and adoption of PRP. Second, we ﬁnd that differences between ﬁrms, ranked by productivity, are related to workforce characteristics. As expected, over-achiever ﬁrms have more trained employees and make less use of ﬁxed-term contracts; they also have lower percentages of women on the staff. Third, we ascertain the unambiguous role of industrial relations: overperformers show high levels of unionization and operate in conditions of higher labor market tightness, but also have fewer recruitment problems in ﬁlling vacant jobs. Lastly, we observe that the ranking order of companies is probably inﬂuenced by other features, such as speciﬁc internal conditions (size and sector) and external factors (geographical location). One clear result is that the greatest incidence of high performers is to be found among large-sized ﬁrms. This result is consistent with that obtained by Pagano and Schivardi (2003), who show that size (because of its inﬂuence on innovation activity) impacts on growth and plays a role in cross-country comparisons for European countries, including Italy. In addition, Table 1 shows diverging patterns between industrial and services sectors: the majority of under-achiever ﬁrms are in trade, hotels and restaurants, whereas the best performers are in manufacturing, conﬁrming the hypothesis that sector specialization contributes to explaining disparities between successful and unsuccessful Italian ﬁrms. Table 1 also shows regional gaps: the highest productivity gains are recorded in North-West Italy, whereas North-Eastern regions do not show clear characterization in terms of company success. Other summary statistics that distinguish ﬁrm with and without PRP are reported in Table 2. Some fundamental differences emerge between ﬁrms adopting PRP and others. The former are generally much larger ﬁrms, and, over the period 2002–2005, achieved higher average value added and higher capital per worker, and were more present in international markets. In terms of industrial relations, we observe a more diffuse presence of unions in PRP ﬁrms, suggesting the existence in those ﬁrms of more intensive collective actions by employees.

179

Performance-Related Pay, Unions, and Productivity in Italy

Table 2.

Statistics by Performance-Related Pay – 2005.

Variable

PRP Average (log) value added 2002–2005 Average (log) ﬁxed capital 2002–2005 Union % Women % Managers % White collars % Blue collars % trained % Fixed-term contracts Foreign Unemployment rate Vacancy rate Firm size No. of employees o10 No. of employees W9 and o50 No. of employees W49 and o250 No. of employees W249 and more Macro-region North-West North-East Centre South Sector Mining, distribution of gas, and others Manufacturing Construction Trade, hotels, and restaurants Transport and communications Financial intermediation Other business services Education, health, and other business service No. of Observations

PRP

No PRP

Whole Sample

Mean

SD

Mean

SD

Mean

SD

10.32 9.60 0.85 0.27 0.03 0.47 0.46 0.25 0.07 0.68 5.44 0.02

1.27 1.73 0.36 0.22 0.05 0.29 0.29 0.32 0.11 0.47 2.90 0.07

8.48 7.37 0.07 0.47 0.04 0.45 0.46 0.12 0.14 0.29 7.36 0.17

1.02 1.72 0.25 0.34 0.11 0.36 0.35 0.27 0.23 0.45 4.76 0.37

0.04 8.60 7.52 0.09 0.46 0.04 0.45 0.46 0.12 0.14 0.31 7.29 0.16

0.19 1.14 1.80 0.29 0.34 0.11 0.35 0.35 0.27 0.22 0.46 4.72 0.36

0.06 0.34 0.46 0.14

0.25 0.48 0.50 0.34

0.60 0.36 0.03 0.00

0.49 0.48 0.18 0.06

0.58 0.36 0.05 0.01

0.49 0.48 0.21 0.09

0.49 0.22 0.19 0.10

0.50 0.42 0.39 0.29

0.39 0.11 0.27 0.23

0.49 0.31 0.45 0.42

0.40 0.11 0.27 0.22

0.49 0.31 0.44 0.42

0.03 0.69 0.03 0.11 0.05 0.00 0.06 0.03

0.18 0.46 0.18 0.31 0.21 0.05 0.23 0.18

0.01 0.26 0.16 0.29 0.04 0.01 0.20 0.04

0.07 0.44 0.36 0.45 0.20 0.11 0.40 0.20

0.01 0.27 0.15 0.28 0.04 0.01 0.19 0.04

0.08 0.45 0.36 0.45 0.20 0.11 0.39 0.20

785

5,375

6,160

Notes: ‘‘Low’’ is the group of ﬁrms whose performances over the period 2002–2005 are minor of the 25th percentile of the average value added per employee distribution, ‘‘Medium’’ indicates the group with past performances between the 25th and 75th percentiles, and ‘‘High,’’ the group with past performances over the 75th percentile.

180

MIRELLA DAMIANI AND ANDREA RICCI

As regards the workforce, the main differences are the higher proportions of men and trained employees, as beneﬁciaries of PRP, and a lower percentage of ﬁxed-term contracts. The Econometric Framework The econometric strategy used to analyze the relation between PRP and ﬁrm productivity is based on a two-step estimation procedure and quantile regression methods. The two-step estimation procedure is similar to that used by Black and Lynch (2001) and allows us to exploit the speciﬁc structure of the AIDA-RIL merged sample, where each ﬁrm presents longitudinal information on balance-sheet variables for the period 2002–2005 and cross-sectional information on PRP adoption and other workplace characteristics for 2005. In the ﬁrst step, panel data methods are used to estimate the parameters of the timevariant input factors (i.e., capital and labor) of a classical Cobb–Douglas production function. OLS and quantile regression methods are used in the second step to estimate the coefﬁcients associated to PRP and other workplace characteristics (obtained from the 2005 RIL survey) over the whole distribution of the ﬁrm-speciﬁc ﬁxed effects estimated in the ﬁrst step. The coefﬁcients of the production function can be estimated consistently with the ﬁxed-effect estimator. However, the within-estimator tends to go too far in discarding potentially valuable cross-sectional information, because the impact of observed (almost) time-invariant factors, such as sector, PRP, and other quasi-ﬁxed variables in the production function cannot be identiﬁed, or measurement errors may explain a large part of their variance (Dearden, Reed & Van Reenen, 2000; Ichniowski, Shaw, & Prennushi, 1997; Griliches & Mairasse, 1995). This feature proves to be a crucial hindrance in our case, because we know only whether or not an establishment adopts PRP in 2005, and this institutional variable does not change much over time. As PRP is treated as a quasi-ﬁxed variable, we assume that ﬁrms which adopted PRP in 2005 also adopted a PRP scheme before and after that year. On the contrary, the use of the (within) ﬁxed effect method could act as the lower-bound estimate of the correlation between PRP and ﬁrm productivity, taking into account ﬁrm unobserved heterogeneity. As shown by the related econometric literature on productivity, recently surveyed by Van Biesebroeck, 2007, estimations of a Cobb–Douglas production function by OLS produces overestimates, due to the correlation between inputs and productivity. Conversely, the ﬁxed effects model gives underestimates, because of limited within group variation (especially in short panels) and

Performance-Related Pay, Unions, and Productivity in Italy

181

ampliﬁes measurement errors. In between these two methods are all the others. In our case, since our main aim is to estimate the correlation between PRP and ﬁrm productivity, taking into account ﬁrm unobserved heterogeneity, we choose to estimate the Eq. (1) using only the within ﬁxed effect method. On the contrary, the use of the GMM method to address potential endogeneity of input variables of the Eq. (1) is not feasible in our case, given the unavailability of longitudinal data for a longer span of time. Note also that the GMM estimator does not allow us to solve the endogeneity problem of the adoption of PRP in 2005. As regards the second step, the quantile regression method offers notable advantages over the least-squares method when the coefﬁcient of PRP varies signiﬁcantly across the distribution of ﬁrm productivity. Such variations occur when, for example, ﬁrm unobserved heterogeneities in terms of management quality and norms of industrial relations map the observed distribution of ﬁrm productivity. Furthermore, quantile estimates are robust relative to least squares estimates when there is signiﬁcant heterogeneity, because they assign less weight to outliers and are robust to departures from normality. We follow the two-step procedure, as described below. In the ﬁrst step, we estimate a classical Cobb–Douglas production function with two input factors, capital and labor. The following speciﬁcation is thus used: ln Y it ¼ a ln K it þ b ln Lit þ T t þ ai þ it

for i ¼ 1; . . . ; 6160; t ¼ 2002; . . . ; 2005

(1) where Yit is the value added of ﬁrm i at time t, Kit the physical capital, Lit the number of employees, Tt the year dummies, to control for the business cycle, ai the unobserved ﬁrm-speciﬁc ﬁxed effect, and eit the idiosyncratic error term. Then Eq. (1) is estimated with a within-estimator. On the basis of these ﬁrst-step estimates, for each ﬁrm we calculate the average estimated ﬁxed effect ai over the period 2002–2005, to obtain an estimate of the establishment-speciﬁc ﬁxed component. The distribution of this component is then regressed over the establishment variables, which are quasi time-invariant, and over the employment characteristics. The second-step estimation is performed with quantile regressions over the following equation: a^i ¼ dy PRPi þ ay U i þ Zy X i þ ui

(2)

where i ¼ 1; . . . ; N is the number of observations in 2005, y the y-quantile being analyzed, a^i the estimate of ﬁrm unobserved ﬁxed effects obtained

182

MIRELLA DAMIANI AND ANDREA RICCI

from the ﬁrst step (Eq. (1)), PRP and U are dummy variables indicating, respectively, the presence of PRP and unions at ﬁrm level, and X is a vector of other control variables. The vector of coefﬁcients dy, a0y , and Zy are estimated at the selected quantile; the idiosyncratic error term, ui, is such that Qy(ui|PRP, U, X) ¼ 0.5 In our framework, coefﬁcients dy and a0y capture the quantile treatment effect of PRP and union presence across the distribution of the estimated ﬁrm speciﬁc unobserved heterogeneity. By verifying whether estimated coefﬁcients d0y a0y differ across the quantile distribution of a^i , we can then infer how these labor market institutions affect ﬁrms’ unobserved ﬁxed component. It is worth noting that the ﬁxed effects, estimated in the ﬁrst step, can be interpreted as average ﬁrm-speciﬁc differences in productivity predicted on the basis of variable inputs – in other words, as the TFP. The estimated ﬁxed effects for the period 2002–2005 indicate whether ﬁrms’ TFP was below or above the average of the other ﬁrms during the observation period. Hence, with quantile regressions, we can infer whether for the key variables we obtain different coefﬁcients across the distribution of the average TFP. The other control variables included in the baseline speciﬁcation of Eq. (2) are the same used for descriptive statistics and includes variables that are standard in the related literature (see e.g., Bastos et al. 2009; Gu¨rtzgen, 2009). Estimations control for worker and ﬁrm heterogeneity, that is for a number of other potential determinants of productivity including the socioeconomic characteristics of an individual (such as gender), and membership three occupational groups (managers and supervisors, white- and bluecollars). We also control whether an individual is employed on a ﬁxed term contract and is trained. The hypothesis is that the heterogeneity of workers (differentiated by gender, tenure, and skills) will inﬂuence the relationships we are testing. The group of ﬁrm characteristics includes size (interpreted, as suggested by Gu¨rtzgen, 2009, as capturing some part of unobserved ﬁrm quality and technological differences) and a group of variables related to strategies: capital accumulation, internationalization, innovation. Internationalization, proxied by a dummy variable indicating whether the ﬁrm sells its products abroad, controls for pressures on ﬁrm strategies exerted by international competition (Melitz, 2003). Other controls, such as local unemployment and vacancy rates, capture the role of labor market tightness and outside options that are expected to inﬂuence employee effort, as predicted by efﬁciency wage models. Estimations also control for unobserved sector characteristics and all regressions include eight industry-dummies corresponding to the

Performance-Related Pay, Unions, and Productivity in Italy

183

manufacturing and services sectors, deﬁned at the two-digit level. In addition, the regressions include four dummies for geographical macroareas to take into territorial disparities in industrial relations, which are very important in the Italian economy.

ESTIMATION RESULTS This section presents the main econometric results. It should be emphasized that our two-step approach does not consider any endogeneity at the second-step level: ﬁrms’ decisions to adopt PRP and the presence of unions may be related to productivity performance. The occurrence of such reverse causality may generate biased estimates of quantile regressions, and we interpret the econometric results as simple correlations between the unexplained part of productivity and labor market institutions. Auxiliary estimations of the determinants of the adoption of PRP schemes, obtained by using a probit model, offer additional insights and are reported in the Appendix (Table A2).6 The estimates of the marginal effects show that gains in productivity in preceding years increase the probability of adoption of PRP, and these marginal effects are higher for the group of high performers. Internationalization, unionization, training, and gender composition are signiﬁcant predictors of the probability of PRP, thus conﬁrming that the two groups of ﬁrms (PRP and non-PRP) are very different.7 As mentioned earlier, to estimate the PRP–productivity relationship, the ﬁrst step consists in estimating Eq. (1) for the unbalanced panel of ﬁrms sampled for the period 2002–2005 with a within-estimator. In the second step, quantile regressions were applied to Eq. (2) to estimate the coefﬁcients for PRP and unions at different points of the ﬁrm-speciﬁc ﬁxed effect distribution, that is, at the 0.10th, 0.25th, 0.50th, 0.75th, and 0.90th quantiles. Columns 1–2 of Table 3 report the estimated coefﬁcients of the production function (1). The ﬁrst-step estimates of the ﬁxed capital and number of employees have the expected signs and high statistical signiﬁcance. We will not comment further on these results, as our focus is on the OLS and quantile estimates for the second-step regressions (columns 3–11 of Table 3). The principal result is that the estimate of PRP is positive and statistically signiﬁcant (error level of 1%) across the whole distribution, in line with the results obtained for the Italian metal-working sector (Biagioli & Curatolo, 1999;

23,058 8,604

No. of observations No. of ﬁrms Pseudo R-squared 6,060 0.592

1.13* 0.20* 0.26* 0.34* 0.13 0.18*** 0.06** 0.43* 0.09* 0.01* 0.56* 0.47* 0.40* 0.37* 0.59* 1.53* 2.31* (Yes)

Coeff.

OLS

0.12 0.03 0.03 0.04 0.10 0.10 0.03 0.07 0.02 0.00 0.10 0.05 0.05 0.05 0.02 0.03 0.05

SE

SE

0.3162

2.73* 0.24 0.17* 0.04 0.24* 0.04 * 0.48 0.06 0.93* 0.20 0.92* 0.20 0.13** 0.04 0.78* 0.14 0.12* 0.04 0.01 0.01 1.05* 0.23 0.45* 0.11 0.44* 0.11 0.34* 0.09 * 0.56 0.05 1.45* 0.06 2.14* 0.10 (Yes)

Coeff.

ql0

0.3566

1.6* 0.19* 0.27* 0.4* 0.20*** 0.16 0.04 0.5* 0.06*** 0.01 0.6* 0.39* 0.33* 0.28* 0.56* 1.41* 2.28* (Yes)

Coeff.

q25

0.13 0.03 0.03 0.03 0.12 0.12 0.03 0.08 0.02 0.00 0.14 0.06 0.06 0.05 0.03 0.04 0.07

SE

q50

0.4051

1.13* 0.20* 0.25* 0.35* 0.19*** 0.22** 0.09* 0.28* 0.10* 0.01* 0.45* 0.48* 0.41* 0.36* 0.56* 1.48* 2.39* (Yes)

Coeff.

Second Step

0.13 0.03 0.03 0.03 0.10 0.10 0.03 0.06 0.02 0.01 0.11 0.05 0.05 0.05 0.02 0.04 0.04

SE

0.4326

0.29*** 0.17* 0.30* 0.26* 0.62* 0.70* 0.09** 0.21* 0.08* 0.01*** 0.35* 0.38* 0.31* 0.27* 0.63* 1.58* 2.46* (Yes)

Coeff.

q75

0.18 0.04 0.03 0.05 0.16 0.15 0.04 0.08 0.03 0.01 0.12 0.07 0.07 0.06 0.03 0.04 0.06

SE

SE

0.4236

0.64** 0.31 0.22* 0.06 0.30* 0.05 * 0.19 0.07 1.30* 0.28 1.37* 0.28 0.04 0.05 0.14 0.13 0.08** 0.04 0.01 0.01 0.34** 0.18 0.35* 0.09 0.23** 0.10 0.29* 0.09 0.65* 0.05 1.61* 0.08 2.41* 0.10 (Yes)

Coeff.

q90

Note: Omitted category: managers, ﬁrms with o10 employees, South.*Signiﬁcant at 1%, **signiﬁcant at 5%, ***signiﬁcant at 10% in quantile regression, bootstrapped errors with 200 replications.

0.12 0.01 0.24* 0.01 (Yes) 7.64* 0.05

SE

Log (ﬁxed capital) Log (employees) Year dummies Constant PRP Union % women % white-collars % blue-collars % trained workers % ﬁxed-term Contracts Foreign Unemployment rate 2004 Vacancy rate North-West North-East Centre 10–49 employees 50–249 employees >250 employees Sector dummies

*

Coeff.

Log (v.a.)

First Step

Labor Productivity and TFP: Two-Step Estimates Using within Estimator in the First Step, OLS, and Quantile Regressions in the Second Step – All Non-Agricultural.

Dependent Variable

Table 3. 184 MIRELLA DAMIANI AND ANDREA RICCI

Performance-Related Pay, Unions, and Productivity in Italy

185

Origo, 2009). We also ﬁnd that the point estimates of PRP are fairly uniform across the distribution, ranging from 0.17 at the 10th and 75th quantiles to 0.22 at the 90th quantile. As mentioned earlier, the estimated coefﬁcients can be interpreted as the partial derivative of the conditional quantile of the estimated ﬁxed effect a^i (i.e., TFP) with respect to a particular regressor, that is qQy(ai|PRP, X)/qX. Then, for instance, the estimated coefﬁcient of 0.17 is the marginal change in a^ i at the 10th conditional quantile due to a discrete change (from 0 to 1) in the adoption of PRP, while the estimated coefﬁcient of 0.22 is the marginal change in a^i at the 90th conditional quantile. A similar interpretation holds for other quantile estimates and other regressors. Furthermore, given that the estimated ﬁxed effects indicate whether ﬁrms’ TFP is below or above the average productivity of the other ﬁrms over the period 2002–2005, the results in Table 3 show the role of PRP in ﬁlling the gap with respect to the productivity average for low achiever ﬁrms or, conversely, in increasing the gap for over achiever ﬁrms. The positive and signiﬁcant estimates of unions are also worthy of note. A feasible explanation is that their presence minimizes free-riding and promotes collaborative attitudes. However, the positive coefﬁcient of the union dummy variable increases signiﬁcantly at the highest quantiles, being 0.24 in the 1st and 0.30 in the 75th and 90th quantiles.8 These results may be explained by arguing, as suggested by Addison and Hirsch (1989, p. 76), that ‘‘union and non-union establishments may differ systematically in the quality of unmeasured organizational factors, so that ﬁrm effects are not independent of union status. For example, inputs such as managerial supervision and the quality of labor relations may be correlated with unionism, and omission of these factors may bias the union coefﬁcient.’’ Table 3 also shows other ﬁndings related to workers’ characteristics, such as employment positions: the coefﬁcient associated to the white- and bluecollar components is negative and signiﬁcant, mainly at the highest quantiles (with respect to the omitted category, i.e., managers and supervisors). A plausible interpretation is that managerial employees have a positive and signiﬁcant inﬂuence on productivity, especially in better-performing ﬁrms. Among other factors, this may be due to their providing better-designed pay schemes to induce optimal effort from their subordinates. As for the other results, the strong negative coefﬁcient associated with the percentage of women (with a decreasing pattern along the distribution, which varies from 0.48 at the 10th quantile to 0.19 at the 90th quantile) should be noted. A cautionary interpretation is necessary, since the percentage of women is very likely to be correlated with unobserved

186

MIRELLA DAMIANI AND ANDREA RICCI

(or omitted) ﬁrm characteristics. The negative coefﬁcient associated with the female component is likely to be related to the gender wage gap, more plausibly at the lowest quantiles of the wage distribution.9 Future research, based on additional data, would certainly help to estimate the impact of the female component and to verify whether lower TFP is caused by less generous PRP offered to women. Another interesting ﬁnding is the negative coefﬁcient of ﬁxed-term workers across the whole distribution, with a magnitude that decreases at higher quantiles and becomes statistically insigniﬁcant at the 90th quantile. Conversely, the estimated coefﬁcient of trained workers is positive, but signiﬁcant at only 1% for the intermediate value of the distribution. Lastly, our estimates conﬁrm that size and the geographical location of ﬁrms are important factors related to their efﬁciency. The positive relationship between ﬁrm size and TFP is signiﬁcant at all quantiles. As expected, we have regional gaps, as shown by the positive coefﬁcients for Northern and Central ﬁrms (with respect to Southern ﬁrms, the omitted category), although the estimates do not show a clear pattern across the distribution.10

MANUFACTURING AND SERVICES SECTORS The role of ﬁrm and sectoral disparities are examined in this section. To ascertain whether the relationship between labor institutions and TFP depends on ﬁrm sectoral specialization, we replicated the preceding analysis for the manufacturing and services sectors, separately.11 The econometric strategy and estimation methods are the same as those used for the whole economy, and do not require further explanation. Table 4 displays the results for manufacturing. The second-step estimates conﬁrm the positive and signiﬁcant coefﬁcient of PRP and unions across all quantiles. However, the magnitudes of the estimated coefﬁcients for both variables are higher than those found for the whole economy, and are fairly uniform across all quantiles. The analysis of the professional composition of the workforce yields no clear interpretation: the coefﬁcients associated with the share of blue- and white-collar workers are positive at the 10th quantile and negative at the 75th quantile, but are not statistically signiﬁcant at the other quantiles. The coefﬁcient for trained workers is positive across the whole distribution, except at the 90th quantile, while that of ﬁxed-term workers is negative and its absolute value is higher at the 75th and 90th quantiles. Estimates for ﬁrm

10,411 3,856

No. of observations No. of ﬁrms Pseudo R-squared

SE

0.17 0.04 0.03 0.05 0.15 0.14 0.04 0.09 0.03 0.01 0.15 0.07 0.07 0.06 0.03 0.05 0.07

2,784 0.6365

1.25* 0.23* 0.26* 0.31* 0.03 0.03 0.13* 0.22** 0.09* 0.01 0.42* 0.32* 0.27* 0.23* 0.52* 1.33* 2.10*

Coeff.

OLS SE

0.33 0.05 0.05 0.10 0.26 0.27 0.04 0.23 0.05 0.01 0.33 0.12 0.14 0.12 0.05 0.07 0.11

0.3162

2.60* 0.22* 0.23* 0.60* 0.76* 0.73* 0.18* 0.35 0.23* 0.01 1.30* 0.37** 0.31** 0.25* 0.53* 1.35* 2.11*

Coeff.

ql0 SE

0.19 0.04 0.03 0.05 0.16 0.15 0.04 0.13 0.03 0.01 0.21 0.07 0.08 0.07 0.04 0.05 0.10

0.3566

1.74* 0.20* 0.27* 0.38* 0.22 0.23 0.11* 0.35* 0.09 0.00 0.57* 0.32* 0.22* 0.17* 0.56* 1.39* 2.34*

Coeff.

q25 SE

0.25 0.03 0.04 0.06 0,17 0.17 0.05 0.10 0.03 0.01 0.17 0.10 0.11 0.09 0.03 0.05 0.08

0.4051

1.29* 0.25* 0.25* 0.32* 0.05 0.08 0.15** 0.02 0.09* 0.00 0.15 0.35* 0.28* 0.22** 0.58* 1.51* 2.42*

Coeff.

q50

Second Step

0.4326

0.68*** 0.21* 0.31* 0.17* 0.34* 0.47* 0.19** 0.16* 0.05* 0.01*** 0.13* 0.29* 0.25* 0.17* 0.64* 1.58* 2.50*

Coeff.

q75

0.25 0.05 0.05 0.06 0.20 0.19 0.05 0.11 0.04 0.01 0.19 0.14 0.14 0.12 0.04 0.06 0.07

SE

q90 SE

0.39 0.08 0.06 0.10 0.36 0.35 0.07 0.19 0.04 0.01 0.24 0.10 0.11 0.10 0.06 0.09 0.12

0.4236

0.02** 0.29* 0.29* 0.10 0.56 0.70** 0.09 0.35** 0.01 0.00 0.08 0.18** 0.12 0.06 0.66* 1.67* 2.59*

Coeff.

Note: Omitted category: managers, ﬁrms with o10 employees, South. * Signiﬁcant at 1%, **signiﬁcant at 5%, ***signiﬁcant at 10% in quantile regression, bootstrapped errors with 200 replications.

0.13 0.01 0.29* 0.01 (Yes) 7.38* 0.08

SE

Log (ﬁxed capital) Log (employees) Year dummies constant PRP Union % women % white-collars % blue-collars % trained % ﬁxed-term Foreign Unemployment rate Vacancy rate North-West North-East Centre 10–49 employees 50–249 employees >250 employees

*

Coeff.

Log (v.a.)

First Step

Labor Productivity and TFP: Two-Step Estimates Using within Estimator in the First Step, OLS and Quantile Regressions in the Second Step – Manufacturing.

Dependent Variable

Table 4. Performance-Related Pay, Unions, and Productivity in Italy 187

188

MIRELLA DAMIANI AND ANDREA RICCI

size and geographical location in industry are similar to those found for the rest of the economy. The estimates for services show high standard errors, probably due to the small number of service ﬁrms using PRP. However, since this is the ﬁrst time that information on this sector of the Italian economy has been available, we have chosen to report our ﬁndings, albeit preliminary. The main difference respect to the manufacturing sector is that in services we have a positive, signiﬁcant, coefﬁcient of PRP only at the 25th and 50th quantiles (Table 5). Our estimates thus suggest that the productivity–PRP link in the services sector is less evident. It is a critical aspect for a laborintensive sector, where promotion of organizational changes and wage policies to motivate employees are expected to enhance TFP growth. The union dummy variable continues to be positive and signiﬁcant across the entire distribution, with a higher coefﬁcient at the 75th and 90th quantiles (around 0.30). Interestingly, trained workers play no signiﬁcant role, whereas estimates for ﬁxed-term workers are negative at all quantiles of the distribution. For over-performer companies, we also ﬁnd that the coefﬁcients of whiteand blue-collars are negative and signiﬁcant (with respect to managerial and supervisory staff, the omitted group). These effects are more noticeable in ﬁrms characterized by low and medium productivity increases. This last ﬁnding also suggests that management counts more in a sector like the tertiary, where production processes are the result of the intangible competences of human capital.12 Lastly, our estimates show that location in Northern regions and size also promote productivity growth in the services sector, as found for manufacturing. One probable reason for the importance of the role of size is that larger ﬁrms are expected to be associated with superior managerial competence. The latter is an omitted variable which could perhaps be proxied by size. Indeed, the best performers (represented by larger ﬁrms) can afford costly strategies such as the upgrading of management: this leads to the implementation of better practices which in turn produces greater efﬁciency. In addition, in large companies, economies of scale reduce implementation costs per employee and explain why beneﬁts are likely to exceed costs. Our results, in any case, call for further investigation of the role of managers in establishing a climate of successful cooperation with workers and their representatives. The analysis of the interactions between PRP and high-performance work practices would allow us to verify the importance of the ‘‘high commitment’’ strategy, based on worker participation and involvement, as found in other studies (see Ichniowski et al., 1997).

SE

12,756 4,748 3,276 0.6365

1.19* 0.12** 0.25* 0.39* 0.24*** 0.28** 0.01 0.53* 0.12* 0.02* 0.66* 0.59* 0.50* 0.48* 0.64* 1.65* 2.42*

Coeff.

OLS

0.16 0.05 0.04 0.04 0.13 0.13 0.04 0.09 0.03 0.01 0.13 0.07 0.08 0.07 0.03 0.05 0.07

SE

SE

0.36 0.08 0.05 0.07 0.28 0.27 0.07 0.21 0.05 0.01 0.30 0.15 0.15 0.13 0.07 0.07 0.13

0.286

2.74* 0.12 0.25* 0.51* 0.98* 0.96* 0.07 1.06* 0.13* 0.00 0.94* 0.39* 0.43* 0.35* 0.61* 1.57* 2.14*

Coeff.

ql0

0.18 0.05 0.05 0.05 0.15 0.15 0.04 0.09 0.03 0.01 0.19 0.08 0.09 0.07 0.04 0.06 0.11

SE

0.3197

1.65* 0.17* 0.24* 0.49* 0.17 0.12 0.02 0.62* 0.03 0.01 0.71* 0.46* 0.40* 0.36* 0.58* 1.53* 2.37*

Coeff.

q25

0.3647

1.16* 0.13* 0.24* 0.36* 0.22*** 0.25*** 0.03 0.45* 0.12* 0.02* 0.58* 0.55* 0.44* 0.41* 0.59* 1.58* 2.53*

Coeff.

q50

Second Step

0.16 0.05 0.04 0.04 0.14 0.14 0.04 0.09 0.03 0.01 0.11 0.07 0.08 0.07 0.03 0.05 0.07

SE

0.3922

0.46*** 0.04 0.31* 0.25* 0.70* 0.77* 0.03 0.26** 0.12* 0.02* 0.53* 0.55* 0.44* 0.43* 0.66* 1.66* 2.51*

Coeff.

q75

0.24 0.05 0.04 0.06 0.20 0.19 0.04 0.11 0.04 0.01 0.17 0.10 0.11 0.09 0.05 0.06 0.08

SE

q90

0.3851

0.65*** 0.08 0.30* 0.24* 1.46* 1.52* 0.02 0.11 0.21* 0.02*** 0.54* 0.56* 0.37* 0.44* 0.67* 1.70* 2.41*

Coeff.

Note: Omitted category: managers, ﬁrms with o10 employees, South. * Signiﬁcant at 1%, **signiﬁcant at 5%, ***signiﬁcant at 10% in quantile regression, bootstrapped errors with 200 replications.

No. of observations No. of ﬁrms Pseudo R-squared

Log (ﬁxed capital) 0.116 0.007 Log (employees) 0.208* 0.009 Year dummies (Yes) Constant 7.768* 0.062 PRP Union % women % white-collars % blue-collars % trained % ﬁxed-term contracts Foreign Unemployment rate 2004 Vacancy rate North-West North-East Centre 10–49 employees 50–249 employees >250 employees

*

Coeff.

Log (v.a.)

First Step

0.37 0.10 0.08 0.09 0.33 0.32 0.07 0.16 0.06 0.01 0.22 0.14 0.14 0.11 0.06 0.09 0.12

SE

Labor Productivity and TFP: Two-Step Estimates Using within Estimator in the First Step, OLS, and Quantile Regressions in the Second Step – Services.

Dependent Variable

Table 5. Performance-Related Pay, Unions, and Productivity in Italy 189

190

MIRELLA DAMIANI AND ANDREA RICCI

CONCLUSIONS Our estimates suggest a positive and signiﬁcant relationship between PRP and productivity in both under and over achiever ﬁrms. Our other main ﬁnding concerns the positive and signiﬁcant role of the presence of trade unions. For this variable, estimates for the whole economy show that signiﬁcant differences exist between ﬁrms: in overachiever ﬁrms, unions, which minimize free-riding behavior and promote cooperative attitudes, have a higher correlation with productivity. Other ﬁndings concern workers’ characteristics. A greater presence of managers is signiﬁcantly related to productivity, especially in ﬁrms with better performances. One probable reason is that the best performers can afford costly strategies, such as the upgrading of management, who then provide better designed pay schemes with the aim of inducing optimal effort from their subordinates. We also ﬁnd a negative relationship between productivity and the percentage of ﬁxed-term workers, while the opposite is true for the proportion of trained workers. Our research also ﬁnds a considerable gap between sectors: higher signiﬁcant estimates of PRP agreements are found for ﬁrms operating in manufacturing industries, uniform along the whole productivity distribution. In the services sector, no signiﬁcant coefﬁcients are obtained for any groups of ﬁrms. It seems likely that the limited implementation of PRP practices or their use as a rent-sharing device may partly explain the slowdown in Italian productivity, mainly due to the bad performance of services. The main limitation of our ﬁndings concerns the possible endogeneity of institutions and human resources management, perhaps due to ﬁrm productivity. Future research will explore this issue by exploiting the second wave of the RIL survey, which allows longitudinal tracking of the adoption of PRP and the presence of unions at ﬁrm level. In this perspective, we will also attempt to evaluate whether employee ﬁnancial participation is a valid strategy, especially in unionized Italian companies. It may be an additional step in detecting the reasons behind the successes and failures of Italian ﬁrms.

NOTES 1. For extensive surveys on the theoretical literature, see Jones and Pliskin (1997), Kruse and Blasi (1997), Pe´rotin and Robinson (1998). 2. These beneﬁcial effects are found also in our case study, Italy, by Jones and Svejnar (1985) who estimate the productivity effects of worker participation in management,

Performance-Related Pay, Unions, and Productivity in Italy

191

proﬁt-sharing, and worker ownership for Italian producer cooperatives, one of the largest systems of producer cooperatives in industrialized Western countries. 3. For a discussion of the experience of employee ﬁnancial participation in Western and Eastern new EU member states, see Uvalic (2006). 4. In particular, we group two-digit sectors into eight categories: (1) mining and quarrying, electricity, gas and water supply; (2) manufacturing; (3) construction; (4) trade, hotels and restaurants; (5) transport and communications; (6) ﬁnancial intermediation; (7) other business services; (8) education, health and other public services. 5. For a detailed discussion on methodological issues and techniques used to perform point and interval inferences, see Buchinsky (1994) and Koenker and Bassett (1978). 6. In our probit estimates, the distribution of average ﬁrm residuals over this period and by different quartiles are used as regressors of the probability of PRP. Note that the vector of control variables does not include the ﬁrm size. This is because in the ﬁrst step we estimate the average ﬁrm residual by controlling for ﬁrm size in the production function (see Eq. (1)). 7. The existence of a virtuous cycle between PRP, unions, and ﬁrms’ productivity has been explicitly veriﬁed by probit estimates, see the Appendix, Table A2. 8. It means that the relationship between unionization and the ﬁrm speciﬁc unobserved heterogeneity of productivity increased from 0.24 to 0.30. Note that the null hypothesis that the estimates are equal between pair-wise quantiles (and across all quantiles) is tested by the variance-covariance matrix of the coefﬁcients of the system of quantile regressions. Consequently, the null hypothesis about the union dummy variable is rejected at the conventional level of signiﬁcance for the 0.10 vs. 0.90 quantiles; conversely, the null hypothesis of equality of the coefﬁcient of the PRP variable is accepted across the entire distribution. 9. The ‘‘sticky ﬂoor’’ hypothesis, i.e., a larger gender pay gap at the bottom of the distribution, has been proposed by Booth (2009) and estimated for Italy (Naticchioni & Ricci, 2009) and 11 European countries, Italy included (Arulampalam, Booth, & Bryan, 2007). 10. Given our focus on institutions, we do not comment on other signiﬁcant estimates of quantile regressions, the positive role exerted by ﬁrms which compete on international markets, and the negative impact of the vacancy rate variable. Both results deserve future study. 11. In particular, the manufacturing sector includes: mining and quarrying; electricity, gas, and water supply; manufacturing; construction. The service sector includes: trade, hotels, and restaurants; transport and communication; ﬁnancial intermediation and other business services; education, health, and other public services. 12. In another study on Italian manufacturing ﬁrms, Piva, Santarelli, and Vivarelli (2005) have shown that organizational improvements, combined with technological innovation, jointly affect the demand for labor and skills, proxied by white- and blue-collar shares.

ACKNOWLEDGMENTS We thank M. Centra, P. Casadio, D. Castellani, G. Chester, A. Matano, M. Morroni, P. Naticchioni, P. Piacentini, E. Rustichelli, M. Uvalic, and

192

MIRELLA DAMIANI AND ANDREA RICCI

F. Lucidi. We also thank participants at the XIII AIEL Conference, University of Brescia, Italy, 2008, and participants at the 15th IAFEP World Conference, Employee Participation for Good Firm Governance, Universite´ Panthe´on-Assas, Paris II, Paris, July 8–10, 2010. All errors are our own.

REFERENCES Addison, J. T., & Hirsch, B. T. (1989). Union effects on productivity, proﬁts, and growth: Has the long run arrived? Journal of Labor Economics, 7(1), 72–105. Bastos, P., Monteiro, N. P., & Straume, O. R. (2009). Firm heterogeneity and wages in unionised labour markets: Theory and evidence. Labour Economics, 16(4), 440–450. Arulampalam, W., Booth, A. L., & Bryan, M. L. (2007). Is there a glass ceiling over Europe? Exploring the gender pay gap across the wages distribution. Industrial and Labor Relations Review, 60(2), 163–186. Bandiera, O., Barankay, I., & Rasul, I. (2005). Social preferences and the response to incentives: Evidence from personnel data. Quarterly Journal of Economics, 120(3), 917–962. Bandiera, O., Barankay, I., & Rasul, I. (2010). Social incentives in the workplace. Review of Economic Studies, 77(2), 417–458. Benabou, R., & Tirole, J. (2003). Intrinsic and extrinsic motivation. The Review of Economic Studies, 70(3), 489–520. Biagioli, M., & Curatolo, S. (1999). Microeconomic determinants and effects of ﬁnancial participation agreements: An empirical analysis of the large Italian ﬁrms of the engineering sector in the eighties and early nineties. Economic Analysis, 2(2), 99–130. Black, S. E., & Lynch, L. M. (2001). How to compete: The impact of workplace practices and information technology on productivity. Review of Economics and Statistics, 83(3), 434–445. Blinder, A. S. (Ed.) (1990). Paying for productivity – A look at the evidence. Washington, DC: The Brookings Institute. Booth, A. L. (2009). Gender and competition. Labour Economics, 16(6), 599–606. Buchinsky, M. (1994). Changes in U.S. wage structure 1963–1987: An application of quantile regression. Econometrica, 62(2), 405–458. Damiani, M., & Ricci, A. (2009). Factors behind performance-related pay: Evidence from Italy. Quaderni di Economia, Finanza e Statistica, Universita` di Perugia, 64. Daveri, F., & Jona-Lasinio, C. (2005). Italy’s decline: Getting the facts right. Giornale degli Economisti e Annali di Economia, 64(4), 365–410. Dearden, L., Reed, H., & Van Reenen, J. (2000). Who gains when workers train? Training and corporate productivity in a panel of British industries. IFS Working Paper 00/01, London. Estrin, S., Grout, P., & Wadhwani, S. (1987). Proﬁt-sharing and employee share ownership. Economic Policy, a European Forum, 2(1), 13–62. Fitzroy, F.R., & Kraft, K. (1992). Forms of proﬁt-sharing and ﬁrm performance: Theoretical Foundations and empirical problems. Kyklos, 45(2), 209–225. FitzRoy, F. R., & Kraft, K. (1987). Cooperation, productivity, and proﬁt sharing. Quarterly Journal of Economics, 102(1), 23–35.

Performance-Related Pay, Unions, and Productivity in Italy

193

Freeman, R., Kruse, D., & Blasi, J. (Eds.) (2010). Worker responses to shirking under shared capitalism (pp. 77–103). Chicago: University of Chicago Press. Godard, J., & Delaney, J. T. (2000). Reﬂections on the ‘‘high performance’’ paradigm’s implications for industrial relations as a ﬁeld. Industrial and Labor Relations Review, 53(3), 482–502. Griliches, Z., & Mairesse, J. (1995). Production functions: The search for identiﬁcation. NBER Working Paper No. 5067. Gu¨rtzgen, N. (2009). Rent-sharing and collective bargaining coverage – Evidence from linked employer-employee data. Scandinavian Journal of Economics, 111(2), 323–349. Holmstrom, B. (1979). Moral hazard and observability. Bell Journal of Economics, 10(1), 74–91. Hu¨bler, O., & Jirjahn, U. (2003). Works councils and collective bargaining in Germany: The impact on productivity and wages. Scottish Journal of Political Economy, 50(4), 471–491. Ichniowski, C., & Shaw, K. (2003). Beyond incentive pay: Insiders’ estimates of the value of complementary human resources management practices. Journal of Economic Perspectives, 17(1), 155–180. Ichniowski, C., Shaw, K., & Prennushi, G. (1997). The effects of human resource management practices on productivity: A study of steel ﬁnishing lines. The American Economic Review, 87(3), 219–313. Jones, D. C., & Pliskin, J. (1991). The effects of worker participation, employee ownership and proﬁt sharing on economic performance: A partial review. In: R. Russell & V. Rus (Eds.), Ownership and participation: International handbook of participation in organisations (Vol. 2, pp. 43–63). Oxford: Oxford University Press. Jones, D. C., & Pliskin, J. (1997). Proﬁt sharing and gain sharing: A review of theory, incidence, and effects. In: D. Lewin, D. J. B. Mitchell and M. A. Zaidi (Eds.), The human resource management handbook, Part 1. Greenwich, CT: JAI Press. Jones, D. C., & Svejnar, J. (1985). Participation, proﬁt sharing, worker ownership and efﬁciency in Italian Producer cooperatives. Economica, 52, 449–465. Kandel, E., & Lazear, E. P. (1992). Peer pressure and partnership. Journal of Political Economy, 100, 801–817. Koenker, R., & Bassett, G. (1978). Regression quantiles. Econometrica, 46(1), 33–50. Kruse, D., Freeman, R., & Blasi, J. (2010). Shared capitalism at work: Employee ownership, proﬁt and gain sharing, and broad-based stock options. Chicago: University of Chicago Press. Kruse, D. L. (1993). Proﬁt sharing does it make a difference? Kalamazoo, W.E. Upjohn Institute for Employment Research, Michigan, USA. Kruse, D. L., & Blasi, J. (1997). Employee ownership, employee attitudes, and ﬁrm performance: A review of the evidence. In: D. Lewin, D. J. B. Mitchell & M. A. Zaidi (Eds.), The human resource management handbook. Greenwich, CT: JAI Press. Part 1. Lazear, E. P. (2000). Performance pay and productivity. American Economic Review, 90(5), 1346–1361. Mas, A., & Moretti, E. (2009). Peers at work. American Economic Review, 99(1), 112–145. Mayer, T., & Ottaviano, G. (2007). The happy few: The internationalisation of European ﬁrms new facts based on ﬁrm-level evidence. Mimeo. Melitz, M. J. (2003). The impact of trade on intra-industry reallocations and aggregate industry productivity. Econometrica, 71(6), 1695–1725. Naticchioni, P., & Ricci, A. (2009). Differenziali salariali di genere: Un’analisi dinamica sulla base delle caratteristiche degli occupati. Rivista delle Politiche Sociali (2), 311–325. OECD. (2004). Employment outlook. Paris: OECD.

194

MIRELLA DAMIANI AND ANDREA RICCI

Origo, F. (2009). Flexible pay, ﬁrm performance and the role of unions. New evidence from Italy. Labour Economics, 16(1), 64–78. Pagano, F., & Schivardi, F. (2003). Firm size distribution and growth. Scandinavian Journal of Economics, 105(2), 255–274. Pe´rotin, V., & Robinson, A. (1998). Proﬁt-sharing and productivity: Evidence from Britain, France, Germany and Italy. Advances in the Economic Analysis of Participatory and Labor-Managed Firms, 6, 135–162. Pe´rotin, V., & Robinson, A. (2003). Employee participation in proﬁt and ownership: A review of the issues and evidence. Working paper, Social Affairs Series, SOCI 109 EN, Luxembourg Parliament, Directorate-General for Research. Piva, M., Santarelli, E., & Vivarelli, M. (2005). The skill bias effect of technological and organisational change: Evidence and policy implications. Research Policy, 34b(2), 141–157. Prendergast, C. (1999). The provision of incentives in ﬁrms. Journal of Economic Literature, 37(3), 7–63. Rusinek, M., & Rycx, F. (2008). Rent-sharing under different bargaining regimes: Evidence from linked employer-employee data. IZA Working Paper No. 3406. Svejnar, J. (1982). On the theory of a participatory ﬁrm. Journal of Economic Theory, 27(2), 313–330. Uvalic, M. (1991). The PEPPER report: Promotion of Employee Participation in Proﬁts and Enterprise Results in the Member States of the European Community, Social Europe Supplement No. 3/91. Luxembourg: Ofﬁce for Ofﬁcial Publications of the European Communities. Uvalic, M. (2006). Employee ownership – Western lessons for the Eastern new member states. In: J. Lowitzsch (Ed.), PEPPER III Report. Dublin, European Foundation for the Improvement of Living and Working Conditions and Institute for Eastern European Studies, Free University of Berlin, pp. 46–68. Van Biesebroeck, J. (2007). Robustness of productivity estimates. Journal of Industrial Economics, 55, 529–569. Wagner, J., Addison, J. T., Claus Schnabel, C., & Schank, T. (2004). Works councils, labor productivity and plant heterogeneity. IZA Discussion Paper No. 1414.

Performance-Related Pay, Unions, and Productivity in Italy

195

APPENDIX Table A1.

Variable Deﬁnition.

Labour productivity Valued added per employee, calculated over the period 2002–2005 (source, AIDA) and deﬂated by the value added deﬂator (source, ISTAT) Fixed capital Fixed capital per employee, calculated over the period 2002–2005 (source, AIDA) and deﬂated by the ﬁxed investment deﬂator at two-digit sectoral level (source, ISTAT) Employees Number of all employees of the ﬁrm PRP Dummy variable that equals 1 if the ﬁrm adopts PRP payments of any kind, 0 otherwise Unions Dummy variable that equals 1 if in the ﬁrm there is a worker representation on any kind, 0 otherwise % women Proportion of women in the ﬁrm % managers/ Proportion of employees of the ﬁrm occupied as supervisors managers/supervisors % white-collars Proportion of employees of the ﬁrm occupied as clerks % blue-collars Proportion of employees of the ﬁrm occupied as manual workers % ﬁxed-term Proportion of employees of the ﬁrm with ﬁxed-term contracts contracts Foreign Dummy variable that equals 1 if the ﬁrm sells abroad 0 otherwise. Unemployment rate Local unemployment rate recorded at the county level of the ﬁrm, 2004 Vacancy rate Number of vacancies in the ﬁrm North-West Dummy variable that equals 1 if the ﬁrm is in NorthWestern regions, 0 otherwise. North Dummy variable that equals 1 if the ﬁrm is in NorthEastern regions, 0 otherwise. Centre Dummy variable that equals 1 if the ﬁrm is in Central regions, 0 otherwise. South Dummy variable that equals 1 if the ﬁrm is in Southern regions, 0 otherwise. Sources: AIDA and ISFOL (RIL Survey for 2005).

196

MIRELLA DAMIANI AND ANDREA RICCI

Table A2.

Probit Estimates of PRP – Year 2005: Marginal Effects.

Variables Estimated TFP: W ¼ 25 and o75 quartiles Estimated TFP: last quartile % women % white-collars % blue-collars % trained % ﬁxed-term contracts Foreign Union Unemployment Vacancy rate North-West North-East Centre Manufacturing Construction Trade, hotels, and restaurants Transportation and communication Financial intermediation Other business services Education, health, and other business serv. Pseudo R2 No. of observations

Whole Economy

Manufacturing

dy/dx

SE

dy/dx

SE

dy/dx

SE

0.02***

0.01

0.03**

0.01

0.01*

0.01

0.13***

0.03

0.17***

0.06

0.06**

0.03

0.03*** 0.01 0.01 0.01*** 0.03* 0.01* 0.21*** 0.00 0.04 0.01 0.07* 0.01 0.00 0.02*** 0.01

0.01 0.03 0.03 0.00 0.02 0.01 0.01 0.00 0.04 0.01 0.04 0.01 0.00 0.01 0.01

0.06*** 0.02 0.01 0.02*** 0.05*** 0.02*** 0.23*** 0.00 0.12*** 0.03*** 0.12*** 0.03***

0.01 0.03 0.03 0.00 0.01 0.01 0.01 0.00 0.03 0.00 0.01 0.00

0.00 0.01 0.01 0.00 0.01 0.00 0.18*** 0.00 0.01 0.02 0.01 0.01

0.01 0.04 0.04 0.01 0.03 0.01 0.03 0.00 0.02 0.03 0.05 0.01

0.00

0.00

0.01 0.02*** 0.01**

0.01 0.01 0.01

0.454

0.3465

3,425

1,388

Services

PART III WORKER COOPERATIVES AND NONPROFIT ORGANIZATIONS

PROFIT REINVESTMENT IN ITALIAN WORKER COOPERATIVES AS A CONTRIBUTION TO A COMMON GOOD: AN EMPIRICAL ANALYSIS ON WORKERS’ PERCEPTION AND MOTIVATION Cecilia Navarra ABSTRACT Italian worker cooperatives display a high proportion of proﬁts reinvested into asset locks: there is some literature investigating their function, but little has been said about workers’ attitude toward them. In this chapter we therefore investigate what workers’ motivations are regarding reinvesting proﬁts into asset locks. We propose to interpret them as a common good and we inquire which factors may increase workers’ willingness to contribute to it. We test two arguments that are provided in the literature on collective action: the effect of having a long-time perspective within the cooperative and the effect of displaying ‘‘collective’’ motivations and preferences other than self-regarding. We perform this test by means of a survey among workers of cooperatives afﬁliated to Legacoop Ravenna in Italy.

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 199–229 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012012

199

200

CECILIA NAVARRA

We identify a positive effect of both factors, although with some distinction. At a ﬁrst glance, we ﬁnd a positive correlation between a longer time horizon and a greater concern for proﬁt reinvestment; when looking closer at the data, we nevertheless see a more complex relationship as two other aspects come into the game: the employment insurance role of worker cooperatives and the ‘‘feeling of belonging’’ that links workers to the ﬁrm. The positive effect of this second aspect on the willingness to reinvest into locked assets is strong, although it only appears among worker-members. Moreover, its effect seem to become greater as workers’ involvement in decision making increases. Keywords: Worker cooperatives; asset locks; common good; workers’ motivations JEL classiﬁcations: J54; J28; D64; D71

INTRODUCTION Italian worker cooperatives display a peculiar feature: a high proportion of proﬁts are reinvested into asset locks, a common fund that is not appropriable and not divisible among members, neither on member’s retirement, nor at the end of the life of the cooperative. Our ﬁrst claim is that asset locks may be qualiﬁed as a ‘‘common pool resource’’ and the contributions to this pool may be interpreted as contributions to a common good. Asset locks are indeed not individually appropriable and they may only be used for the ﬁrm’s overall purposes. These different uses bring about mutual rivalry, but the beneﬁts produced are not excludable. Therefore, we presume that a problem concerning collective action may come up. This may appear in two forms: ﬁrst of all, there may a free rider problem in effort provision on behalf of individuals. Proﬁt reinvestment into such a nondivisible and nonappropriable fund may indeed be inefﬁcient from the incentives point of view. The second problem is an intertemporal one: the members of each generation may be tempted to behave opportunistically with respect to future generations, and opt for a high degree of proﬁt distribution today, as the choice to accumulate will prevent them from having access to those assets tomorrow. Decisions concerning asset locks are, thus, both individual decisions about participation in the creation of a common good through effort provision, and collective decisions regarding

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

201

the allocation of proﬁts to this ‘‘common’’ and their assignment to parts of it for alternative uses. Although a debate ﬂourishes in the literature discussing their function, little has been said about the workers’ perception of asset locks. In this chapter, we intend to investigate what the individual characteristics are that increase the willingness to reinvest proﬁts into the ﬁrm. We investigate the determinants of a positive attitude toward the indivisible accumulation of capital, considering this attitude as a measure of the extent of ‘‘cooperation’’ among members in the collective action. These individual features may be both motivational elements that introduce preferences other than self-regarding, and characteristics concerning the time perspective of each member within the group. First of all, we intend to present the proﬁt allocation mechanism brieﬂy as regards asset locks and a short literature review on their possible functions. Then, we will introduce two arguments that have been put forward in order to explain how cooperation can emerge as a dominant strategy within groups: on one hand, the time horizon of group members and the discount rate of future beneﬁts, taking inspiration from the work of the Nobel laureate Elinor Ostrom (1990), and, on the other, the intervention of some mechanisms of self-recognition of members as part of a ‘‘collective agent’’ (Bowles & Gintis 2006; Hollis & Sugden, 1993). We have strong historical evidence, indeed, that a major role is played by the path of cohesion within the group that forms the cooperative (Menzani, 2007; Nardi, 1998; Navarra, 2010). In the ﬁnal part, we will try to apply these two arguments to the choice of collective investment, by means of an empirical investigation carried out in the workers’ cooperatives associated with the Lega delle Cooperative e Mutue in the province of Ravenna (Italy), using a dataset constructed through an original survey among a sample of workers, integrated with historical evidence and qualitative interviews of the cooperative board members. We will conclude with some reﬂections on the determinants of the positive attitude toward asset locks and on the validity of the aforementioned arguments in our case study. As we will discuss, our ﬁndings are consistent with both arguments, although with some distinctions: the positive opinion on asset locks is nonmonotonic with respect to the time perspective of workers inside the cooperative, and the higher loyalty to the ﬁrms that reinvest a greater share of proﬁts can be interpreted in multiple ways. The positive interplay of the preference for proﬁt reinvestment and the feeling of belonging to the cooperative is strong, but only applies to worker-members. Finally, a possible role for the employment protection function of the cooperative enterprise is highlighted.

202

CECILIA NAVARRA

ASSET LOCKS IN ITALIAN WORKER COOPERATIVES Asset locks (riserve indivisibili) are a common fund that is nondivisible and nonappropriable by members, that is accumulated through the assignment of a yearly proﬁts share that is decided at the end-of-the-year assembly, where all members vote the board’s proposal on the subject. Once proﬁts have been allocated to this fund, they become part of the cooperative’s assets and may only be used for a ﬁrm’s overall purposes, such as new investments and the introduction of innovations. They may be used to recover from losses, but only after having used other divisible funds and with the obligation of their immediate refunding. As regards proﬁt allocation, the limitations imposed by law are to allocate 30% (20% until 2003) of proﬁts to the legal reserve (again, not divisible and not appropriable), and 3% to fondi mutualistici, that are solidarity funds used to support the cooperative movement as a whole. Besides these limitations, proﬁts may be distributed to members in two main ways: as a remuneration of capital stakes (within strict limitations imposed by law), or as a remuneration for the work relationship (the so-called ristorno). Since 2001,1 a second form of distribution has been introduced, that is by increasing individual capital quotas, in order to distribute proﬁts and avoid the risk of undercapitalization of the ﬁrm. Moreover, since 2003,2 cooperatives have been able to accumulate divisible reserves that can be appropriated by the ‘‘ﬁnancial members,’’3 allowing the asset owned by these members to internalize the increases in ﬁrm’s proﬁts, thus becoming more attractive. Up to some years ago we would have argued that asset locks were the only tool for the cooperative enterprise to accumulate capital, but this is no longer the case today. Nevertheless, there is evidence that Italian worker cooperatives reinvest a huge amount of yearly proﬁts into asset locks: a study of the Centro Studi Legacoop (2006) shows that, on average, 86.8% of net proﬁts is reinvested in the ﬁrm and 10.2% is distributed to members. This evidence is conﬁrmed in the literature on the Italian cooperative movement (Bassi 2003, Zevi 2003) and by the data of our case study, as we will illustrate in the following pages. These data show an average share of proﬁt reinvested into asset locks that is clearly above the amount required by law.4 How can we explain this choice? One of the usual arguments to explain this huge reinvestment is the advantage of indivisible accumulation from the point of view of corporate tax. However, we do not think that this is a

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

203

crucial reason: if it is true that there is a ﬁscal advantage as far as asset locks accumulation is concerned this has sharply declined over the last few years,5 but the average share of reinvested proﬁts did not follow this decrease and – at least in our case study – remained almost constant over time (Navarra, 2010). One possible alternative explanation is to assume that asset locks have a collateral function that prevents the ﬁrm’s capital stock from being exposed to huge capital variability (Tortia, 2002): this may explain why weaker ﬁrms choose to reinvest a higher share of proﬁts in asset locks (Jones & Svejnar, 1985) and why the same is done by smaller ﬁrms (Centro Studi Legacoop, 2006). In other works, we focus on another possible argument so to explain the function of asset locks, that relates to the employment insurance role of worker cooperatives: cumulated indivisible capital may be seen as a possible tool for worker members to insure themselves against the risk of losing their jobs and, at the same time, to stabilize wage earnings (Navarra, 2010). Sacconi and Seppi (2006) argue that asset locks (or, rather, reserve indivisibility as a norm) may be interpreted as a governance device for members to ‘‘tie their hand’’ so as not to have the incentive to behave opportunistically toward future members. The authors claim that the indivisibility of the ﬁrm’s assets in cooperatives may be interpreted as a tool to produce the dynamics of the indeﬁnitely iterated game, that is indeed able to make the agents converge on the cooperative equilibrium and avoid free riding (that on the contrary would be a risk if members could withdraw their quota of the ﬁrm’s reserves ad lib.). We do not intend to put forward arguments that conﬂict with the aforementioned; in this chapter we propose another perspective regarding this issue. Although the previous arguments explain asset locks through their function, we have tried to approach the issue from the point of view of workers’ motivation. Moreover, those explanations include a ‘‘common good’’ component: the function of asset locks as a collateral, as an insurance tool, or as a governance device, all include a collective component that is not directly appropriable by individuals. This means that a certain amount of cooperation is needed to achieve all those purposes by means of indivisible accumulation. This is the reason why we investigate what inﬂuences the attitude of workers of cooperatives toward this form of proﬁt reinvestment. We start with a theoretical introduction to those motivational elements that may enhance cooperation among members of a group in the pursuit of a common aim that requires the renouncement to an individual and immediate gain.

204

CECILIA NAVARRA

A THEORETICAL FRAMEWORK ON PROFIT REINVESTMENT AS A COLLECTIVE ACTION: HOW TO ENHANCE COOPERATION If we consider asset locks as a common good, theory tells us that a risk of opportunistic behavior is likely to emerge. As claimed in the seminal work by Mancur Olson (1965), when a group of people share the same interest in order to obtain a good that is, in their perspective, a public good, therefore nonexcludable, there is the risk that they may ‘‘rationally’’ choose not to contribute to its production (each one will not be excluded from the beneﬁts and has no incentive to bear the cost of its provision). If that is so from the part of all members, the common good will not be provided (the so-called risk of ‘‘collective action failure’’). As we have already mentioned, the problem concerning collective action may be twofold: ﬁrst of all, an incentive problem at an individual level in effort provision and, secondly, the risk of opportunism in the current group of members, that may opt for maximising their short-run income by proﬁt sharing, instead of reinvesting proﬁts into asset locks. As regards the ﬁrst, cooperatives differ from the capitalistic ﬁrm, where the risk of free-riding in teams is reduced by the fact that one individual takes control of the activity of the others, as claimed by the ‘‘Property Rights School’’ (Alchian & Demsetz, 1973; Demsetz, 1967; Jensen & Meckling, 1979): he is entitled to monitoring and has efﬁcient production incentives (given by his residual claimancy), whereas the others act on his behalf and under his control. On the contrary, cooperatives are what Elinor Ostrom (1990) would deﬁne as a ‘‘group of principals’’ (Navarra & Tortia, 2010), that is an organization where no one loses his control power and where the ﬁrm’s assets are a common good. What does the literature on collective action say about the factors that may allow a group sharing a common interest to avoid the outcome that is illustrated by Olson?

Time Horizon of Members: Indeﬁnitely Iterated Games and the Long-Term Work Relationship One factor that may allow cooperation to emerge as a dominant strategy within groups is the length of the time horizon of members’ interaction. A longer time horizon, that is the perspective of a stable interaction among

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

205

members in the future, following Elinor Ostrom (1990), allows for reputation mechanisms to arise and the importance given to future beneﬁts to increase. Within the case studies carried out by Ostrom, the more successful groups in common resource management are those marked by the stability of the referring population. The importance given to belonging to the group and to the reputation within it have the effect of lowering the discount rate of the ﬂow of future beneﬁts that derive from the common resource. This mechanism, together with a pattern of credible sanctions, has the effect that group members make long run gains due to participation greater than short run gains that may be obtained from defection (Platteau, 2004). The time horizon of members inside the cooperative is a customary argument claiming the inefﬁciency of this kind of ﬁrm (Furubotn & Pejovich 1970; Jensen & Meckling 1979), as it implies that workers will behave in order to maximize nearer term cash ﬂows6. If we adopt the collective action perspective, the subjective time horizon of members’ interaction may be one of the ways out for the impasse of collective action failure. If we allow for repeated interaction among agents, the importance attached to the possibility of being part of the group tomorrow (and therefore having access to the common good) increases. Within the context of worker cooperatives, this can be translated into two ways. The ﬁrst is proposed by Sacconi and Seppi (2006), who argue that the indivisibility of the cooperative’s assets may be interpreted as a governance device in order to produce the dynamics of the indeﬁnitely iterated game. Therefore, asset locks are an instrument that allows the cooperative to establish the rules to generate a game where it is rational for all the agents to play the cooperative strategy, as they participate in an interaction that potentially repeats itself after each play. This makes the free rider option less rewarding than the choice of continuing to cooperate. As regards capital accumulation, conﬂict among generations may arise, where older members have the tendency to free ride, because of their short-time horizon within the cooperative; thus, if they prevail, the reinvestment of proﬁts will be suboptimal. Conte and Ye (1995) propose a model where different generations have the possibility to enter into a binding agreement that the authors formalize as a system of transfer payments; the outcome of their model is that the cooperative will reach the same steady-state capital stock as the ‘‘twin’’ capitalistic ﬁrm. If we follow Sacconi and Seppi’s suggestion, we may argue that a similar role may be played by asset locks. These can be interpreted as a way for the insiders to decide – as long as they have the incentive to do so – to ‘‘tie their hands’’ referring to the capital they will be tempted to withdraw tomorrow.

206

CECILIA NAVARRA

The second argument relates to the evidence that cooperatives manage to have a more stable workforce than comparable capital-managed ﬁrms (Bartlett, Cable, Estrin, Jones, & Smith, 1992, for example, show that cooperatives display lower quit rates) and guarantee more stable employment relationships. This expands the time horizon of workers inside the ﬁrm, therefore increasing their willingness to reinvest proﬁts into it. This can be even stronger if the incumbent member of the cooperative thinks that his children will work in the same ﬁrm as he does. If we ﬁnd evidence of an intergenerational perspective, it can represent likewise a strong investment incentive and an instrument to avoid free riding toward future members of the cooperative.

The ‘‘We-Rational’’ Attitude of Members: Collective Aims and Feeling of Belonging Major authors claim that interpretations in terms of self-regarding preferences are not sufﬁcient to shed light on many aspects of economic life. Bowles and Gintis (2006) clearly state that they are not satisﬁed with a theory of choice that only allows for reciprocity if interaction takes place in a repeated game framework. They claim that we have to account for the capacity of people to internalize moral norms and to enter into relationships marked by some feeling of belonging. Akerlof and Kranton’s (2005), for instance, analyze how noneconomic concerns (identity) come into economic dynamics: if an agent identiﬁes himself as an insider, he will require lower monetary incentives to do his job. His identity comes from his belonging to a social category, that is not the result of a rational choice, but rather of a ‘‘framing’’ process with regard to ‘‘a person’s self-image as an individual and as a part of a group’’ (Akerlof & Kranton, 2005, p. 1). These approaches therefore raise criticism, as expressed by Hollis and Sugden (1993), about instrumental rationality and strategic interaction models. They argue that these do not leave any ‘‘room for reﬂection’’ between utilities and rational choices: every motivating factor only works as a source of utility and can have no role once utilities are identiﬁed. The attempt would be to give a role to personalized interactions, in order to account for a more complex idea of transactions, and not to scale them down to exchanges (Gui, 2005). This alternative approach that gives new weight to the social context of an action has clear implications in the problem of public good provision. The basic theory on public goods implies that the quantity provided will fall

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

207

short of the efﬁcient level, as each contribution is considered to be a perfect substitute for the others and this implies mutual crowding out of contributions. As Bardsley (2005) points out, we need other approaches to account for the amount of empirical evidence that shows, on the contrary, a crowding-in effect among contributions. He particularly stresses in two concepts that reject mere instrumental rationality: expressive rationality and collective rationality (also Hollis & Sugden, 1993). Expressive rationality implies that an action is not rational in sight of some end (a payoff), but it is rational as it expresses attitudes or principles that have, as regards the agent, the role of realizing his identity. If agents are concerned about the meaning of their actions, not by their consequences, the constraints on actions are not simply given by the speciﬁc current interaction, but by the attitudes prevalent in the community. In this sense, contributing to a public good may be a statement of value and therefore judged irrespectively of the payoff it generates. Collective rationality, in a related way, introduces the possibility for people to agree on what is good for them as a group. It means ﬁrst of all that people can act as collective agents: it becomes rational to act as a part of the group and to take the existence of a ‘‘team-optimal’’ plan as a reason for action. Moreover, in this case, members’ expectations regarding the behavior of others’ are derived from their team membership, not from strategic reasoning. On what concerns worker cooperatives, the observed big amount of proﬁt reinvestment into asset locks represent a direct cost for the individual that is translated into a common good, whose beneﬁts for the individual are strongly diluted within the group of members and over time. We thus introduce the possibility that workers’ motivations go beyond instrumental rationality, in order to account for the fact that individuals may incur net costs, for the beneﬁts of the whole group. The investment into asset locks can be interpreted, on one hand, as deriving from a set of ‘‘normative expectations’’ (Sugden, 1998), that is the mutual expectation to follow a norm and disapprove of breaching it, regardless of the payoff it has for each individual, On the other hand, it can be interpreted as an expression of ‘‘we-rationality,’’ or team-thinking (Hollis & Sugden, 1993), that is, by assuming that what is rational for the individuals is to behave consistently with their belonging to the group. We therefore search for some evidence of these mechanisms, by using our survey data: are there any people who think that it is right to allocate net proﬁts to asset locks and therefore follow this rule, instead of rationally calculating their payoff under different allocation of proﬁts? Another question to be raised is whether there is a form of strategic interaction

208

CECILIA NAVARRA

between the workers and the ﬁrm. The idea is to understand whether the worker ‘‘punishes’’ the cooperative by opposing proﬁt reinvestment into asset locks, in case the ﬁrm does not match his expectations. Last of all, we investigate the existence of a feeling of belonging to the cooperative and the effect that this may have on workers’ attitude toward the indivisibility of ﬁrm’s assets and the lack of proﬁt distribution. It has to be said that the possible function of the cooperative enterprise as an actor of social cohesion and solidarity is documented in the historical literature. The role of cooperatives as solidarity networks has been a major issue since the beginning of the cooperative movement, particularly in the case of Romagna, where the ﬁrst groups to be formed were made up of the weakest categories of workers: as Zangheri, Galasso, and Castronovo (1987) underline, the central feature of the ﬁrst land-workers’ cooperatives was mainly the capacity of solidarity, rather than economic performance. An interesting factor that is widely underlined by historians is the capacity of the cooperative to provide collective goods for the community: holidays for members’ children, activities for the elderly, social insurance mechanisms such as internal pension funds to integrate retired members’ incomes, and cultural activities. These solidarity actions were likely to be sustained even in hard times when the members themselves might have received lower wages (Tampieri, 1967).

INDIVIDUAL CHARACTERISTICS OF MEMBERS AND THEIR OPINION ON ASSET LOCKS: THE EMPIRICAL ANALYSIS The Survey: Lega delle Cooperative e Mutue of the Province of Ravenna The survey was conducted in 2007 by the author within 18 of the 60 worker cooperatives afﬁliated to the Legacoop in the province of Ravenna, Italy. Questionnaires were submitted to workers (both members and non members) for a total of 415 observations. The sector repartition is the following: 87 questionnaires from agriculture, 95 from construction, 52 from manufacturing, 52 from porters’, and 129 from other services. Around 80% of the samples are worker-members, whereas 20% are employees; 40% women; and 66% blue collar. Qualitative semistructured interviews were conducted in the same ﬁrms together with the president or a board member.

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

209

The survey data is matched with ﬁrm-level data based on balance sheets accounting. There is a long history of cooperative movement in the Ravenna region (Romagna). At the end of the nineteenth century the ﬁrst cooperatives of day land-workers were funded, closely followed by the construction ones. Employment protection was a major concern (Sapelli & Zan, 1971), and it is still, even today (Navarra, 2010). The population of Legacoop Ravenna worker cooperatives conﬁrms the high tendency to proﬁt reinvestment into asset locks displayed by the Italian cooperative movement: the average share of yearly proﬁts accumulated into asset locks is nearly 88%, and more than 60% of them reinvest between 90% and 100% of proﬁts into the indivisible fund.

The Time Horizon Perceived by Members and Their Attitude Toward Asset Locks Our measure of members’ perception of the duration of their relationship with the cooperative includes both the individual perspective and the intergenerational one. In the ﬁrst case, we look at the time that has already been spent in the cooperative and at the subjective perception of people about their permanence in it. The second one has two components. On the one hand, the backward-looking perspective: did any of the respondent’s family work in the cooperative before he did? On the other hand, the forwardlooking one: does the respondent expect that his children will work in the same cooperative as he does? As regards the individual time dimension, cooperative workers have the perspective of long permanence within the ﬁrm. The average time spent in the cooperative among the respondents is around 13 years, but may vary in different sectors.7 If we look at the frequency of contract typologies, we can ﬁnd a very skewed frequency table, where 70% of interviewees have a nonﬁxed term employment contract.8 The subjective perception of the future time permanence in the ﬁrm provides striking evidence, as may be seen in Table 1. Only 14% of the interviewees think they will leave in about 5 years time, whereas 72% think they will stay at the cooperative until they retire. This is a very mixed indicator as it captures workers’ satisfaction, age, employment stability, and outside options at the same time. It gives nevertheless a ﬂavour perception of stability within the cooperative. As we have already argued, the sectors where the life-long employment pattern is more diffused are

210

CECILIA NAVARRA

Table 1.

Time Perspective in the Cooperative (Frequency Table).

Up to 2 years Between 2 and 5 years More than 5 years Until I retire

5.79% 8.31% 13.60% 72.29%

‘‘How long do you plan keep on staying in your current workplace?’’

agriculture and construction, where about 80% of respondents declare they expect to stay for their whole working life. If we look at the intergenerational perspective, the backward-looking dimension is captured by a dummy variable (‘‘family’’ dummy), that has a value of 1 if the respondent has at least one member of his family who worked at the same cooperative before he did (a little more than 26% of respondents).9 The third component is the forward-looking one: a dummy variable (‘‘children’’) is generated from two questions, where workers were asked if they would like their children to work at their cooperative and whether they think that it is likely to happen: 19.8% of respondents expect their children to work in the cooperative.10 The time horizon is therefore long in terms of a single generation, but it does not extend to future generations11 to the same extent. Concerning the difference among members and employees,12 we notice that the latter show an average permanence in the ﬁrm (6 years) that is lower than the former (15 years); they also display a lower perception of future time horizon, but keep the same pattern as members: if among members 78% of respondents think they will stay at the cooperative until they retire, this percentage decreases to 51% among employees but as regards this question it is still the option they select the most. When asked if they would move to a capitalistic enterprise, more respondents among employees, with respect to members answered that they would if they were offered a higher wage or a better position, and less of them refused a priori this option (17.5% among employees and 33% among members). This is consistent with the information obtained through qualitative interviews with board members, where it was often underlined that it is usually a choice of the worker himself not to become member, as it implies both a monetary investment in the cooperative, and a choice of long-term relationship with the ﬁrm.13 We now inquire the relationship between the time horizon of workers and the attitude to collective accumulation of capital, that is the choice of

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

211

renouncing to an individual use of proﬁts today, in favour to a collective use tomorrow. Concerning the opinion on collective accumulation in the questionnaire, workers are directly asked to give their opinion on asset locks (see Table 2, column a). It is clear that a positive opinion is prevalent, ﬁrst of all motivated by the ﬁrm’s solidity, secondly by a moral concern regarding fairness. Workers were also asked to indicate their preferred destination of proﬁts (Table 3, column a). Two answers are dominant: proﬁts shall be kept in the cooperative in order to strengthen it, and proﬁts shall be distributed to members, but only in good times. On looking at the whole picture, the opinion regarding proﬁt reinvestment and asset locks accumulation seems to show a time horizon effect in both questions: if we split our observations among those workers who have a time horizon lower than 5 years within the company, and those who have a higher one, we can see that 66% of workers in the ﬁrst group have a positive opinion regarding the indivisible accumulation of capital and 39% believe that proﬁts should be kept within the ﬁrm. In the second group, these percentages rise to 76% and 46%. Nevertheless, if we present the answers in a more detailed way, the picture becomes more mixed. If we look at the opinion table on asset locks, disentangled by the time expected to be spent in the cooperative, we will ﬁnd some mixed evidence, as we can see in Table 2 columns (b–e). Among those who think they will stay in the ﬁrm until they retire, more than 80% of respondents believe that asset locks are either necessary or fair. Furthermore, a similar distribution can be found among those whose time perspective is between 2 and 5 years as well. Moreover, 55% of respondents in column b (those who think they will stay at the cooperative for less than 2 years) say that asset locks are necessary in order to strengthen the enterprise: they consider the solidity of the cooperative important even though they are not likely to have any advantages, given the short run projection they have in the cooperative. A more straightforward picture emerges if we consider how people answered the question relative to the preferred destination of proﬁts (Table 3, column b–e): among the ‘‘short-time-horizon’’ respondents, proﬁt distribution is preferred on the whole. The argument concerning the ﬁrm’s strengthening becomes more frequent when we look at the ‘‘long-timehorizon’’ members, making proﬁt reinvestment becomes more attractive: workers who think that they will have a longer relationship with the cooperative are more concerned with its solidity.

12.9 35.48 9.68

30

55 5

9.8

41.18

23.53

25.49

8.02

44.66

6.49

40.84

3.9

44.16

6.49

45.45

9.54

44.08

12.17

34.21

3

33

8

56

10.32

48.04

12.1

29.54

‘‘Do you think that cooperative’s asset locks arey’’ (a) whole sample of respondents, from (b) to (e) observations split by time perspective in the ﬁrm, (f) subsample of respondents who expect their children to work in the same cooperative, (g) subsample of respondents who don’t expect their children to work in the same cooperative, (h) subsample of respondents who had a relative working in the same cooperative before him/her, (i) subsample of respondents who had not a relative working in the same cooperative before him/her.’’

41.94

10

Whole Time Time Perspective Time Time Perspective Children: Children: Relatives: Relatives: Sample Perspective between 2 and 5 Perspective until retirement Yes (f) No (g) Yes (h) No (i) (a) (%) o2 years years (c) (%) W5 years (d) (e) (%) (%) (%) (%) (%) (b) (%) (%)

Fair (also toward future 36.48 generations) Unfair (members should 11.02 appropriate of proﬁts) Necessary for coops’ 44.09 solidity A tool for board members 8.4 to keep the wealth inside the ﬁrm

Opinion on AL

Table 2. Opinion on Asset Locks (Frequency Table).

212 CECILIA NAVARRA

36.84

15.79

26.32

21.05

19

34.04

14.78

32.19

35.48

22.58

22.58

19.35

37.25

7.84

45.1

9.8

30.65

13.79

35.63

19.92

33.97

9.09

38.96

18.18

31.79

16.23

32.78

19.21

24.51

12.75

44.12

18.63

35.02

15.52

30.32

19.13

Whole Time Time Time Time Perspective Children: Children: Relatives: Relatives: Sample (a) Perspective Perspective Perspective until Retirement Yes No (g) (%) Yes (h) (%) No (i) between 2 (%) o2 years W5 years (e) (%) (f) (%) (%) (b) (%) and 5 years (c) (d) (%) (%)

Opinion on Proﬁt Destination (Frequency Table).

‘‘What do you think shall be done with ﬁrm’s proﬁts?’’

Shall be distributed to members Shall be accumulated into the ﬁrm to strengthen it Shall be accumulated into the ﬁrm because it’s of everyone and not an individual wealth Shall be distributed to members but only in good times

Opinion on Proﬁt Destination

Table 3.

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations 213

214

CECILIA NAVARRA

If we look at the effect of the intergenerational link, we ﬁnd out that this increases the importance given to fairness. The positive opinion on indivisible proﬁt reinvestment as they are ‘‘fair’’ (including the concern for future generations) has a greater weight if the respondent expects his own children to work in the same ﬁrm as he does. The same result can be found by disentangling these opinions by the ‘‘family dummy.’’ Here, the increased importance given to ‘‘fairness’’ regarding the indivisibility of asset locks is even higher. This is conﬁrmed by qualitative evidence: the cumulated assets by reinvesting proﬁts are considered to be the inheritance that has been left by previous generations and this is the reason why it has to be indivisible.14 When asked what should be done with ﬁrm proﬁts, the ‘‘solidity argument’’ is more relevant among those who have a longer projection in the enterprise: looking at the answers relevant to the question disentangled by ‘‘children dummy,’’ we can see that the two groups display similar patterns, but the respondents who think (or hope) that their children will work in the same ﬁrm as them have a greater concern for the solidity role of proﬁt reinvestment in the enterprise. If we try to understand whether the accumulation policy of the ﬁrm inﬂuences members’ time horizon, the evidence is mixed. By looking at the relationship between the current share of proﬁts reinvested into asset locks and the percentage of respondents that have a long-time horizon within the ﬁrm (or project this on future generations) on the total number of questionnaires collected in each cooperative, we do not ﬁnd any evidence of correlation; if we use an indicator of the cumulated (instead of current) asset locks, there is evidence of a positive correlation, but only in highly capitalized ﬁrms.15 A more accurate speciﬁcation may be obtained by looking at the determinants of the number of years that workers have actually spent in the cooperative, where we can ﬁnd a positive effect of the share of reinvested proﬁts in the common fund (Table 4). Workers tend to stay longer in cooperatives that reinvest a higher share of proﬁts into asset locks: this may be due to the fact that asset locks actually work as an incentive device in order to increase the willingness of workers to stay. This result is not unambiguous, it can also be explained by the fact that ﬁrms that reinvest a greater share of proﬁts are more able to guarantee workers employment stability. Moreover, we ﬁnd a mixed effect depending on the size of the cooperative: if measured by the number of workers, it has a negative effect (bigger cooperatives may have higher turnover), but the opposite is true if we

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

Table 4.

215

Speciﬁcation for the Number of Years Already Spent in the Cooperative (OLS Estimates).

Number of years already spent in the cooperative ¼ y Share of proﬁts to asset locks(mean over t) Sector dummy construction Sector dummy manufacturing Sector dummy services Sector dummy porters Workers (mean over t) Members’ share capital (mean over t) Age of respondent Education of respondent Is the respondent worker member? (dummy) Is the respondent a woman? (dummy) Did the respondent have a relative working in the same coop? (dummy) Has he/she worked in a KMF before? (dummy) Const R2

0.671 0.479 (n.s.) 0.9 (n.s.) –0.371 (n.s.) –0.55 (n.s.) –0.104 0.00000084 0. 63 –1.0 3.7 0.07 (n.s.) 2.88 –1.5 –17.4 0.615

Signiﬁcant at the 90% level, signiﬁcant at the 95% level, and signiﬁcant at the 99% level.

measure the size of the cooperative by the amount of share capital. Besides some intuitive results (the positive effect of age and membership), it is interesting to notice that education displays a negative coefﬁcient, that may be due to the fact that more educated workers are usually younger and have more outside options. Having a relative working in the same cooperative increases the years spent inside it: the intergenerational projection on the cooperative seems to be a loyalty factor and possibly an element of ‘‘belonging.’’ To sum up, we may say that we have evidence of higher loyalty in ﬁrms that accumulate a greater share of proﬁts in asset locks: however, it is difﬁcult to disentangle whether it is a ‘‘time horizon’’ effect, or it is more likely to be due to an employment insurance mechanism that works through asset locks (Navarra, 2010). If we look at the time perspective perceived by interviewees, unambiguous evidence of the effect of the accumulation policy as a device to ‘‘extend’’ the time horizon in ﬁrms as perceived by workers is missing. It is obvious that this does not rule out the possibility that indivisibility as a norm may be interpreted as a device to avoid opportunism as far as future generations are concerned (Sacconi & Seppi, 2006). On the contrary, we ﬁnd some evidence of a positive effect of workers’ time horizon on their opinion on proﬁt reinvestment and asset locks. If we consider the

216

CECILIA NAVARRA

opinion expressed about the destination of the cooperative’s yearly proﬁts, workers who have a longer time perspective seem to prefer proﬁt accumulation as it represents a tool to reinforce the ﬁrm. Thus, if there is a longer permanence perspective in the cooperative, there may be a lower discount rate of future beneﬁts deriving from the interaction with the ﬁrm, that partially translates into the willingness to reinvest proﬁts into asset locks. If, however, we look at the opinion directly expressed on asset locks, there is no clear evidence of a monotonic effect of the time perspective, since we may ﬁnd a positive opinion among members who plan to quit the ﬁrm relatively soon. There seem, nevertheless, to be a greater concern for fairness among those who have an intergenerational projection. This evidence calls for a closer analysis of the existence of some feeling of belonging to the cooperative.

The Belonging to a ‘‘System of Relationships’’ As we presented, two questions in the survey deal with workers’ opinion on asset locks and on what, according to them, should be done with the ﬁrm’s yearly proﬁts (see Tables 2 and 3). Using these two answers, we have constructed a dummy variable that aims to capture those who have answered in favour of asset locks both in the ﬁrst and in the second, in order to have an indicator of the ‘‘collectivist type’’ of worker.16 The outcome is that 40% of respondents have been deﬁned as ‘‘collectivist.’’ Does this ‘‘type’’ of worker exhibit speciﬁc characteristics? The intention is to work out what the factors are that favour a positive attitude toward collective accumulation. First of all, there is a difference where the perception of outside options is concerned. While in the noncollectivist type the dominant answer to the question ‘‘if you were asked to, would you move to a capitalistic enterprise?’’ is ‘‘only if I were offered a higher wage’’ (53%), the collectivist type displays a higher subjective loyalty to the ﬁrm (this option is chosen by 39.6% of respondents), and a higher frequency of the answer ‘‘no, in any case’’ (among ‘‘non-collectivist’’ this was the answer of 23.7% of the sample, that rises to 39% among ‘‘collectivist’’ workers). Another interesting issue is how they look upon their relationship with the ﬁrm. The ‘‘collectivist’s’’ answer displays a remarkable decrease in the frequency of the answer ‘‘a normal working relationship’’ with respect to ‘‘noncollectivist’s’’ (from 47.6% to 29.3%), and an increase in the frequency of the answers ‘‘a contribution to the creation of better working conditions’’

217

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

(from 4.9% to 13.4%) and a ‘‘system of relationships that goes beyond the job’’ (from 20.3% to 32.9%). Some differences are also shown regarding what the priority of the cooperative should be. Although job stability is the most frequent answer in both groups (about half of both subsamples), the ‘‘collectivist’’ group seems to give more importance to the quality of working conditions: the answer ‘‘the priority of the cooperative shall be the provision for better working conditions than elsewhere’’ was chosen by 27.6% of the ‘‘noncollectivist’’ and by 41.2% of the ‘‘collectivist’’ workers. Therefore, the importance given to the on-the-job consumption and to the relational aspects of work, seems to increase in the group that is in favour of indivisible accumulation. Before turning to the determinants of the ‘‘collectivist’’ attitude, we intend to go through the deﬁnition of some other related variables, starting with a variable that captures whether the cooperative matches workers’ expectations: do workers react by ‘‘punishing’’ or ‘‘rewarding’’ the ﬁrm if it does or does not match their expectations? The idea is to see whether workers display a strategic interaction with the cooperative or if they choose proﬁt destination ‘‘unconditionally.’’ As a measure for matching workers’ expectations and ﬁrm’s behaviour, we have used two questions, where respondents are asked ﬁrst of all what the cooperative’s main concern shall be and secondly what the priority actually is (see Table 5).17

Table 5.

Priority of the Cooperative (Frequency Table). What Shall be the Main Concern What is the Main Concern of the of the Cooperative? (a) (%) Cooperrative in Reality? (b) (%)

Share out as much proﬁts as possible to members Provide a stable and secure job Provide job in better conditions than elsewhere Maximize proﬁts Other

13.12

10.26

46.53

39.49

33.17

25.38

5.2 1.98

21.79 3.08

(a) ‘‘What shall be the main concern of the coop?’’ and (b) ‘‘What is the main concern of the cooperative in reality?’’

218

CECILIA NAVARRA

First of all, we have to notice a ‘‘public good’’ preference of the interviewees: around 80% declare that job stability or the quality of working conditions are their priorities. We consider ‘‘disappointed’’ those respondents who choose one of these two and found that the aim of the cooperative was not so: they are 18.8% of the sample. We consider a positive opinion on reinvestment of proﬁts into asset locks as rewarding for the ﬁrm (both as a sign of trust and as sign of what we deﬁned a ‘‘collectivist’’ attitude), whereas we consider the opposite to represent a ‘‘punishing’’ attitude. As we can see in Table 6, the ‘‘disappointed’’ do not have a more hostile attitude toward asset locks; we can even see that greater importance is given to them as an instrument for a ﬁrm’s solidity, and that there is lower frequency of the opinion that they are just a tool so that board members do not share ﬁrm’s revenues. Albeit knowing the limitations of our proxies, we can argue that, even if worker members choose to reinvest proﬁts as they expect some beneﬁts from this choice, this does not seem to strictly depend on the fact that a set of expectations are achieved. This evidence is geared toward limiting the weight of strategic reasoning in individual choices and allows for the introduction of some elements of ‘‘normative expectations,’’ as discussed in the previous section. Last of all, we try and ﬁgure out a way to measure the feeling of belonging to the cooperative; we ﬁnd it in the question where the workers were asked to deﬁne their relationship with the cooperative (Table 7). Although 40% of the respondents do not identify any difference between their relationship with the cooperative and a normal job, we consider that the second, third, and ﬁfth answers indicate a feeling of belonging to the cooperative, that includes the relational aspect and the idea of a common aim. Following this, we constructed a dummy variable indicating whether

Table 6.

Opinion on Asset Locks, by Dummy ‘‘Disappointed’’.

Opinion on AL

Fair (also toward future generations) Unfair (members should appropriate proﬁts) Necessary for coops’ solidity A tool for board members to keep the wealth inside the ﬁrm

Disappointed (%)

Not-Disappointed (%)

35.71 11.43 48.57 4.29

36.66 10.93 43.09 9.32

219

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

Table 7.

Relationship with the Cooperative (Frequency Table).

a normal work relationship a network that goes beyond the work relationship a contribution to better working conditions A way to get protection on the job An engagement for common objectives

40.24% 25.37% 8.29% 4.88% 21.22%

‘‘How would you deﬁne your relationship with the cooperative?’’

Table 8.

Opinion on Asset Locks and on Proﬁt Destination, by DUMMY ‘‘belonging’’.

Opinion on AL

Fair (also toward future generations) Unfair (members should appropriate proﬁts) Necessary for coops’ solidity A tool for board members to keep the wealth inside the ﬁrm Opinion on proﬁt destination. ‘‘Proﬁts shall bey’’ Distributed to members Accumulated into the ﬁrm to strengthen it Accumulated into the ﬁrm because it’s of everyone and not an individual wealth Distributed to members, but only in good times

Belonging Not-belonging (%) (%) 42.33 6.98 48.84 1.86

28.92 16.27 37.95 16.87

12.04 41.2 13.43

28.22 24.54 16.56

33.33

30.67

the worker gives a ‘‘team’’ meaning to its job relationship: 54.22% of respondents display a so-deﬁned feeling of belonging to the cooperative.18 The types of workers identiﬁed by this variable show signiﬁcant differences in their opinion on asset locks. As may be seen in Table 8, both the consideration of them as ‘‘fair’’ and as ‘‘necessary’’ increase among this type of workers. Moreover, there is a sharp decrease in the option chosen by those who think that asset locks are nothing but a tool for board members to keep revenues inside the ﬁrm; this can be a sign of greater trust in the cooperative and in its governing body. Also the opinion about how proﬁts should be allocated displays a different pattern if the respondent reveals a feeling of belonging: among ‘‘belonging’’ respondents, there is a sharp decline in the frequency of those who think that proﬁts should be shared out to members in any case. Those

220

CECILIA NAVARRA

who prefer accumulation for insurance reasons is higher. This may indicate that the concern for a ﬁrm’s stability and the importance given to employment protection are interlinked with the feeling of belonging and they inﬂuence the opinion on collective accumulation. The ‘‘team thinking’’ aspect can in fact be both cause and effect of the attempt to defend better working conditions and greater employment stability. This is the main conclusion that can be achieved by running a regression in order to look for the determinants of the ‘‘collectivist’’ type. We use a PROBIT model with the dummy ‘‘collectivist’’ as a dependent variable (Table 9, column a). Both the share of proﬁts reinvested into asset locks and the per capita cumulated amount of locked assets of the cooperative display a positive effect regarding the probability that a worker will have a positive opinion on them. This result is interesting as it can show the feedback effect of the collective choice on members’ opinion. The fact itself of working in a cooperative where asset locks play an important role shapes the preferences of workers, in favour of a proﬁt reinvestment policy.19 The dummy variable indicating the feeling of belonging has a strongly signiﬁcant positive effect: this conﬁrms the previous evidence and displays a high correlation between the positive opinion on reserve indivisibility and the importance of the relational component on the job. We ﬁnd also evidence of a positive effect of the variable that indicates whether the respondent feels protected both from employment and wage ﬂuctuations (‘‘insured’’); this may conﬁrm the aforementioned interplay between ‘‘collectivist’’ attitude, feeling of belonging and job security. The size of the cooperative has a negative effect on the probability that the worker has a collectivist attitude, both if measured by the number of workers and as the amount of share capital: this can be interpreted as a – broadly speaking – collective action problem. The relationship between individual involvement and the ﬁrm’s policy is weaker in bigger cooperatives. It can be also a consequence of a greater degree of managerialization, that the decision regarding the allocation of proﬁts is taken by a smaller group of people with respect to the size of the entire workforce. The variable indicating whether the worker is ‘‘disappointed’’ with respect to the aims of the cooperative does not display an effect that is signiﬁcantly different from zero, consistently with what has been said beforehand. The negative coefﬁcient of the dummy variable that indicates whether the respondent is a member and not simply an employee is worth looking at: a

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

221

Table 9. Speciﬁcation of the Variable ‘‘Collectivist’’ (PROBIT Estimates, Marginal Effects). Whole Sample of Respondents, Subsample of Worker Members, Subsample of Non Member Employees. Dummy ‘‘Collectivist’’ ¼ y

Marginal effects Whole Sample

Members

Employees

Share of proﬁts to asset locks (mean 0.0073 0.006 (n.s.) –0.10 (n.s.) over t) 0.39 (n.s.) –0.999 Sector dummy construction 0.52 Sector dummy manufacturing –0.487 –0.42 0.34 (n.s.) Sector dummy services 0.11 (n.s.) –0.05 (n.s.) –0.957 Sector dummy porters 0.40 (n.s.) 0.24 (n.s.) –0.96 Workers (mean over t) –0.000648 –0.0003 (n.s.) 0.0005 (n.s.) Cumulated asset locks per capita 0.000014 0.00013 –0.00002 (n.s.) (mean over t) Share capital (mean over t) –0.0000000435 0.0000000043 0.000000028(n.s.) Is the respondent worker member? –0.277 – – (dummy) –0.375 –0.1935 (n.s.) Is the respondent a blue collar? –0.296 (dummy) –0.93 (n.s.) –0.268 Has he/she worked in a KMF before? –0.158 (dummy) Did the respondent have a relative 0.13 (n.s.) 0.17 –0.07 (n.s.) working in the same coop? (dummy) Age of respondent 0.001 (n.s.) – 0.0053 (n.s.) –0.0002 (n.s.) Education of respondent –0.014 (n.s.) –0. 062 (n.s.) 0.19 Is the respondent ‘‘insured’’? (dummy) 0.136 0. 183 –0.07 (n.s.) Is the respondent ‘‘belonging’’? 0.25 0.25 –0.07 (n.s.) (dummy) Is the respondent ‘‘hightime’’? (dummy) –0.04 (n.s.) –0.04 (n.s.) 0.05 (n.s.) Is the respondent ‘‘disappointed’’? 0.11 (n.s.) 0.14 (n.s.) 0.04 (n.s.) (dummy) – Did the respondent ever participate to the – 0.24 decision on end-of-the-year proﬁt allocation? Pseudo-R2 0.30 0.36 0.29 signiﬁcant at the 90% level, signiﬁcant at the 95% level, AND signiﬁcant at the 99%

level.

possible explanation is that employees do not expect to beneﬁt from the’ distribution of any possible proﬁts and therefore are more favourable toward proﬁt accumulation. We get a more detailed insight by looking at the answers to the question about their opinion on proﬁt allocation,

222

CECILIA NAVARRA

disentangled by membership status: only 7.8% of employees thinks that proﬁts should be shared out to members, whereas among members the proportion is 21.9%; 33.8% of employees think that proﬁts should be kept within the ﬁrm as it is fair (‘‘it belongs to everyone and it is not individual wealth’’), while among members this proportion is less than 10%. On the contrary, the opinion on asset locks is quite similar among members and nonmember employees. If we split the observations among members and employees and perform the same regression as we did previously in order to look for the determinants regarding the probability that a worker will be ‘‘collectivist,’’ we see that the subsample of members almost reﬂects the results obtained on whole sample. On the contrary, among non members, the explanatory variables that were signiﬁcant among members cease to be so: the positive feedback between a favourable attitude toward collective accumulation, the feeling of belonging to the cooperative and the employment protection it guarantees, only seem to work for worker members. An important point that can be highlighted only among members is the role of participation at the end-of-the-year decision on proﬁt allocation20: the fact of participating in the decision makes the worker-member more willing to reinvest proﬁts into asset locks.

CONCLUDING REMARKS The aim of this chapter is to investigate a peculiar feature of Italian worker cooperatives, the high propensity to reinvest proﬁts into asset locks, that is a common fund nondivisible and nonappropriable by individual members. In the literature, we have found some arguments that explain its possible functions. Here, we intend to interpret it as a common pool resource and therefore to use the tools provided by the literature on collective action in order to get a different insight into the issue. Our question is what inﬂuences workers’ motivations in order to enhance their willingness to reinvest proﬁts in the common fund. In the literature, we have found two main arguments to explain how the suboptimal free rider equilibrium may be avoided, by enhancing cooperation when contributing to a common good: repeated interaction among members and assumptions on rationality other than instrumental ones (namely ‘‘collective rationality’’). By means of a survey among a sample of workers of cooperatives afﬁliated to the Legacoop in the province of Ravenna (Italy), we have investigated to what extent these arguments are relevant to understanding the attitude of workers toward asset locks.

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

223

First of all, we have evaluated the weight of the ‘‘time horizon’’ argument, that is the long-term relationship which workers set up within the cooperative. Time horizon is measured both in an individual and in an intergenerational perspective: if there is evidence that a large number of people tend to think that they will stay in the cooperative for their whole working life, not many of them expect their children to work in the same ﬁrm as they do. The relationship between the time projection within the ﬁrm and the opinion on asset locks and proﬁt destination is not unambiguous. Concerning the effect of the accumulation policy of the cooperative on members’ permanence in it, we ﬁnd some evidence of stronger loyalty to ﬁrms that accumulate a greater share of proﬁts into asset locks, but we cannot disentangle whether it is due to the creation of a repeated interaction mechanism, or it is more likely due to the employment insurance that may be provided by the cooperative by means of asset locks accumulation. If we look at the incidence of the ‘‘time horizon’’ on the choice of proﬁt destination, we ﬁnd the expected effect (longer time horizon associated to a greater interest in the solidity of the ﬁrm) when workers are asked what they think about proﬁt destination. If they are asked about indivisibility of the cooperative’s assets, the relationship is non monotonic, as many respondents who have a short-time perspective are concerned with their ﬁrm’s solidity likewise those who think they will stay in the cooperative for more time. Secondly, what we ﬁnd is an increasing concern for ‘‘fairness’’ (asset locks as a ‘‘fair’’ allocation of the gain obtained by the collective effort) among those who have an intergenerational projection, both ‘‘forward-looking’’ and ‘‘backward-looking.’’ These last two observations tell us that there is room for assuming a positive role of ‘‘collective rationality,’’ as we do in the second part of the chapter. We then investigate the self-recognition by members as a part of a common project: this aims to go beyond the idea that people act following merely instrumental rationality, and to consider that people may ‘‘rationally’’ act as part of a collective agent. We have shown that workers identiﬁed as ‘‘collectivists’’ (those having a positive attitude toward asset locks) display some peculiar characteristics, namely a preference for the relational aspects of the job and for ‘‘on-the-job’’ consumption. We have then analyzed what the determinants of this attitude are. The main ﬁnding is a positive effect of the variable capturing worker’s feeling of belonging to the cooperative: our hypothesis of a role played by ‘‘collective rationality’’ seems to be conﬁrmed. Secondly, we ﬁnd a positive effect of the variable

224

CECILIA NAVARRA

indicating whether the worker feels insured with respect to employment and wage ﬂuctuations: we argue that a positive relation exists among teamthinking, a positive attitude toward asset locks, and the role of the cooperative in protecting employment. We have not found any ‘‘punishing’’ strategies from workers who are disappointed with the satisfaction of the ﬁrm’s purposes, that shows workers’ nonstrictly conditional behaviour toward the cooperative. Interestingly, members who participate in the decision on proﬁt allocation are more willing to reinvest proﬁts: this can be an indication for board members, that may induce them to favour participation at members’ assemblies, if they aim to obtain a more favourable attitude to asset locks. Some difference can be found among worker members and employees: evidence tells us that the latter are on average more favourable to asset locks rather than the former. This can be explained by the fact that only members participate in proﬁt sharing. If we split the observations in the two subgroups, we ﬁnd that the positive interplay between the ‘‘collectivist’’ attitude and feeling of belonging disappears among those employees that are not members of the cooperative. We have learned through this case study that both the time perspective of workers in the cooperative and their feeling of belonging play a role in shaping preferences on proﬁt destination. At a ﬁrst glance, we ﬁnd a positive correlation between a longer time horizon and a greater concern for proﬁt reinvestment, that is the expected effect. Looking closer at the data, we nevertheless see a more complex relationship as two other aspects come into the game: the employment insurance role of worker cooperatives, and the ‘‘feeling of belonging’’ that links workers to the ﬁrm. This is clear if we look at the positive relationship between the intergenerational projection in the cooperative and the concern for fairness when supporting the nonappropriability of the ﬁrm’s assets. On the contrary, the positive effect of ‘‘team thinking’’ on the willingness to reinvest into locked assets is very strong, although it only appears among worker-members. The motivations among nonmembers seem to be different given that their opportunity cost of proﬁt reinvestment into asset locks is lower. More may be done to discuss proﬁt destination in worker cooperatives and the underlying mechanisms. We have tried to apply some reasoning in this work that may appear far from the literature on cooperatives: the results are geared toward taking into greater account the ties between the worker and the ﬁrm that are different to the simple job relationship.

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

225

Moreover, the effect of these ties seems to become greater as workers’ involvement in decision making increases.

NOTES 1. Law 142/2001concerning worker members in cooperatives. 2. Company Law Reform. 3. These are some members that only have a ﬁnancial stake in the ﬁrm and own assets that are designed to help the cooperatives to ﬁnd capital as a substitute to the stock market. We have evidence that these tools are underutilized in Italian worker cooperatives (Bonfante, 2008; Zevi, 2005), or are used to keep retired members in the cooperative (from qualitative interview to board members in the present case study). 4. The legal reserve, that is 30% of yearly proﬁts. 5. Since 1977, revenues accumulated into asset locks were completely absolved of corporate tax, up to 2002, when a reform reduced the tax-free share of proﬁts allocated in such fund to about 53%. In 2004, this decrease was mitigated, but the law stated that 30% of cooperative proﬁts would be taxed anyway regardless of its destination. If the reason for indivisible accumulation were the reduction in taxes, we would thus expect a consequent reduction in the share of reinvested proﬁts. 6. The ‘‘horizon problem’’ is a well-known argument that claims the inefﬁciency of the cooperative enterprise, as it foresees constant underinvestment by workermembers. Following this argument, workers lack incentives to invest in the ﬁrm because they have a truncated time horizon, that is their employment perspective in the ﬁrm: they therefore appraise investments with respect to their employment in the cooperative, that means that they maximise near term net cash ﬂows, while they will not undertake those project whose payoffs occur far in the future. 7. The lower permanence is shown in the service sector (7.8 years), where there are the youngest cooperatives; if they are not so young, such as in the cleaning services, they are more likely to have a higher turnover. The longest permanence is in the agricultural sector (18.9 years), followed by the construction one (15.8 years): these are the sectors where coop have historically played the role of ensuring life-long employment. 8. In this respect, the above-average sectors are construction, manufacturing, and porters; agriculture displays that 55% of respondents have a ﬁxed-term contract, but generally speaking this is due to the fact that they are used in applying this type of contract and to iterate it for a de facto nonﬁxed term relationship (source: author’s interview to board members). The sector that displays a more varied spectrum of typologies (19% ﬁxed-term contracts and 14% collaboration agreements). 9. The picture varies depending on the sector: agriculture has the highest share (54%). This is conﬁrmed by interviews with board members, even though they declare that this tendency is declining among new members. 10. We consider the respondent to ‘‘expect that his children will work in the same coop’’ if he answers that his children are already working at the same ﬁrm as he is, if he thinks that it is likely to happen, and if he would simply like it to happen. The

226

CECILIA NAVARRA

reason for including the last category is that we want to identify those workers that subjectively have an intergenerational perspective and how this inﬂuences their behaviour within the ﬁrm. 11. The same data disentangled by sector provides a different picture to the ‘‘backward-looking’’ intergenerational link: the decline is greater in agriculture and among porters, whereas as far as the construction sector is concerned almost the same proportion of people who had a relative working in the coop before him declared they expected their children to work there; last of all an increase can be seen in the service sector. 12. Note that employees are on average younger (36 years compared to 44 members), more educated (almost 40% have a BA that only 9% have among members) and have a higher qualiﬁcation on the job (41% among them are blue collar, while this percentage is 72% among members). 13. It happens quite often that workers enter the cooperative as employees and then opt out for membership: the advantage for the cooperative in having more members than simple employees is twofold. On one hand, every member raises capital in the cooperative and, on the other hand, cooperatives have a ﬁscal advantage as they are considered as‘‘a mutualita` prevalente’’, a category introduced in 2003 that identiﬁes the cooperatives that mostly operate with members from those that do not. In the case of workers cooperatives, to enter into this category, at least half of the labour costs of the coop has to be ascribed to worker-members. 14. Source: author’s interview to board members. 15. To be more precise, in cooperatives with cumulated asset locks per capita that is greater than 40,000 euros. If we perform a PROBIT model with y ¼ the respondent has an ‘‘open’’ time horizon, we do not ﬁnd a signiﬁcant effect of the asset locks measures. As it is intuitive, the age of a respondent and the fact of being member positively affects his/her time horizon within the coop (Navarra, 2008). Moreover, the correlation disappears if we use an intergenerational time perspective indicator. 16. We want to identify those workers who are in favour of asset locks accumulation consistently throughout the questionnaire, since we have seen some possible incoherent answers in the two speciﬁc questions. 17. In both questions they have the same possible answers. 18. As for the collectivist type, the attitude toward outside options is different following the ‘‘feeling of belonging’’. Although for the ‘‘not-belonging’’ agents, the dominant answer is that they would move to a capitalistic ﬁrm if they were offered a higher wage for the ‘‘belonging type’’ this is no longer a clear preference for ‘‘the belonging type’’: if we compare ‘‘not-belonging’’ with ‘‘belonging’’ agents, this answer declines from 55.74% to 40.57%, whereas the frequency of those workers who say that they would never move to a capitalistic ﬁrm rises from 20.77% to 37.74%. 19. It may be argued that a reverse causality problem may be in progress: nevertheless, the number of the interviewees in the sample is too small to think of a causality that goes from the fact of being a ‘‘collectivist’’ to the ﬁnal decision of the ﬁrm’s policy. 20. The obvious reason is that nonmembers do not take part in a ﬁrm’s decisions.

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

227

ACKNOWLEDGMENTS I’m indebted ﬁrst of all to Professor Enrico Luzzati, who was professor at the University of Turin and has been to me an important scientiﬁc guide. I would also like to thank Dott. Ermanno Tortia and Professor Carlo Borzaga from University of Trento and EURICSE and Professor Viriginie Pe´rotin (University of Leeds). On what concerns the empirical analysis, I am strongly indebted to Circolo Cooperatori Ravennati and Legacoop Ravenna (ﬁrst of all to Mario Tampieri and Stefano Patrizi) and to Dott. Sergio Giaccaria (University of Torino). I am also indebted to an anonymous referee for his precious comments. The whole responsibility of any error or omission is of the author alone.

REFERENCES Akerlof, G. A., & Kranton, R. (2005). Identity and the economics of organizations. Journal of Economic Perspectives, 19(1), 9–32. Alchian, A. A., & Demsetz, H. (1973). The property right paradigm. Journal of Economic History, 33(1), 16–27. Bardsley, N. (2005). Interpersonal interaction and economic theory: The case of public goods. In: B. Gui & R. Sugden (Eds.), Economic and social interaction: Accounting for interpersonal relations. Cambridge, UK: Cambridge University Press. Bartlett, W., Cable, J., Estrin, S., Jones, D., & Smith, S. (1992). Labor-managed cooperatives and private ﬁrms in north-central Italy: An empirical comparison. Industrial and Labor Relations Review, 46(1), 103–118. Bassi, A. (2003). Le cooperative tra dimensioni dell’impresa e guadagni dei soci: il problema dei ristorni nel pensiero degli economisti moderni. In: B. Jossa & V. Buonocore (Eds.), Le organizzazioni economiche non capitalistiche: economia e diritto. Bologna, Italia: il Mulino. Bonfante, G. (2008). Societa` cooperativa. In: Il Diritto, Enciclopedia Giuridica del Sole 24 Ore. Milano, Italy: Il Sole 24 Ore. Bowles, S., & Gintis, H. (2006). The evolutionary origins of collective action. In: B. R. Weingast & D. A. Wittman (Eds.), Oxford handbook of political economy. Oxford, UK: Oxford University Press. Centro Studi Legacoop. (2006). Distribuzione del valore aggiunto: Cooperative e loro stakeholders. In: A. Zevi & P. Petrucci (Eds.), Attivita` e Ricerche del Centro Studi LegaCoop. Rome, Italy. www.cslegacoop.coop Conte, M.A., & Ye, M-H. (1995). An overlapping generation model of investment in labormanaged ﬁrms. In: D. Jones & J. Svejnar (Eds.), Advances in the economic analysis of participatory and labor-managed ﬁrms (Vol. 5, pp. 159–174). Amsterdam, Netherlands: Elsevier. Demsetz, H. (1967). Towards a theory of property rights. American Economic Review, 57(2), 347–359.

228

CECILIA NAVARRA

Furubotn, E., & Pejovich, S. (1970). Property rights and the behaviour of the ﬁrm in a socialist state. Journal of Economics, 30(3–4), 431–454. Gui, B. (2005). From transactions to encounters: The joint generation of relational goods and conventional values. In: B. Gui & R. Sugden (Eds.), Economic and social interaction: Accounting for interpersonal relations. Cambridge, UK: Cambridge University Press. Hollis, M., & Sugden, R. (1993). Rationality in action. Mind (New Series), 102(405), 1–35. Jensen, M., & Meckling, W. (1979). Rights and production functions: An application to labormanaged ﬁrms and codetermination. Journal of Business, 52(4), 469–506. Jones, D., & Svejnar, J. (1985). Participation, proﬁt sharing, worker ownership and efﬁciency in Italian producer cooperatives. Economica, 52, 449–465. Menzani, T. (2007). La cooperazione in Emilia Romagna. Dalla Resietenza alla svolta degli anni ‘70. Bologna, Italia: il Mulino. Nardi, S. (1998). Il lavoro. Alcune considerazioni. In: Circolo Cooperatori Ravennati. La memoria ritrovata. Fonti orali e storia della cooperazione ravennate. Atti dell’incontro di studio del 9 maggio 1998. Ravenna, Italia: Longo Editore. Navarra, C. (2008). Collective action, employment insurance, we-rationality: An empirical investigation on Italian worker cooperatives. PhD dissertation, University of Turin, Italy. Navarra, C. (2010). Collective accumulation of capital in Italian worker cooperatives between employment insurance and ‘‘we-rationality’’: An empirical investigation. Euricse Working papers no. 004/10. Trento, Italy. Navarra C., & Tortia E. (2010). Employer’s moral hazard and the emergence of worker cooperatives. Paper presented at the XV IAFEP Conference at Universite´ Panthe´onAssas Paris II, July 2010. Olson, M. (1965). The logic of collective action. Public goods and the theory of the groups. Cambridge MA: Harvard University Press. Ostrom, E. (1990). Governing the commons. The evolution of institutions for collective action. Cambridge, UK: Cambridge University Press. Platteau, J. P. (2004). Solidarity norms and Institutions in Agrarian societies: Static and dynamic considerations. In: S. Kolm & J. Mercier-Ythier (Eds.), Handbook of gift-giving, reciprocity and altruism. Amsterdam, Netherlands: North Holland and Elsevier. Sacconi, L., & Seppi, M. (2006). Le basi etiche e economiche della promozione cooperativa, classica e non classica. In: M. Bulgarelli & M. Viviani (Eds.), La promozione cooperativa. Bologna, Italia: il Mulino. Sapelli, G., & Zan, A. (1971). Costruire l’impresa: la CMC di Ravenna dal 1945 al 1972. Bologna, Italia: il Mulino. Sugden, R. (1998). Normative expectations: The simultaneous evolution of institutions and norms. In: A. Ben-Ner & L. Putterman (Eds.), Values and organizations. Cambridge, UK: Cambridge University Press. Tampieri, M. (1967). I problemi del socio nelle cooperative di produzione e lavoro. Atti del Convegno Provinciale – Ravenna, 11/3/1967. In: La Cooperazione Ravennate, 16, pp. 39–46. Tortia, E. (2002). The internal organisation of labour managed ﬁrms: the problem of value added distribution and of capital accumulation. PhD dissertation, Universita` di Ferrara e Bologna.

Proﬁt Reinvestment in Asset Locks: Workers’ Motivations

229

Zangheri, R., Galasso, G., & Castronovo, V. (1987). Storia del movimento cooperativo in Italia 1886–1986. Torino, Italia: Einaudi. Zevi, A. (2003). Il fenomeno cooperativo e le modiﬁche in corso nell’ordinamento italiano. In: B. Jossa & V. Buonocore (Eds.), Le organizzazioni economiche non capitalistiche: economia e diritto. Bologna, Italia: il Mulino. Zevi, A. (2005). Il ﬁnanziamento delle cooperative. In: E. Mazzoli & S. Zamagni (Eds.), Verso una nuova teoria della cooperazione. Bologna, Italia: il Mulino.

DOES TRAINING POLICY HELP TO ATTRACT, RETAIN, AND DEVELOP VALUABLE HUMAN RESOURCES? ANALYSIS FROM THE MONDRAGON CASE$ Imanol Basterretxea and Eneka Albizu ABSTRACT The aim of this chapter is to ascertain the degree to which a training policy developed through corporate training centers is recognized as a source of competitive advantage for attracting, developing, and retaining valuable staff. The ﬁeldwork is based on a survey of Human Resource (HR) managers from 66 cooperatives of the Spanish Mondragon cooperative group. The empirical test carried out conﬁrms that Mondragon’s training policy, backed up by its corporate training centers, is perceived by HR managers as a tool that provides advantages to attract, develop, and retain valuable human resources. The results also suggest that those advantages are more moderate than has been cited in classic literature on Mondragon. The results of this study can be helpful for the growing

$

Some of the results of this work were presented in the 2nd International CIRIEC Research Conference on Social Economy. October 1–2, 2009, O¨stersund Ja¨mtland (Sweden).

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 231–260 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012013

231

232

I. BASTERRETXEA AND E. ALBIZU

number of companies choosing to create and reinforce corporate training centers. The link between training policy and the perceived ability to attract and retain valuable employees showed in this case can also be helpful for other companies that, as Mondragon, face limitations in wage policy. This chapter contributes to the literature on the educational fabric of Mondragon adding updated empirical evidence and incorporating the point of view of HR managers of the group’s cooperatives. With respect to the contribution of this chapter to the literature on training policy, the chapter’s ﬁndings, in particular those regarding the effect of training on worker attraction and retention, add empirical evidence to the few studies on the subject. Keywords: Training; employee attraction; employee retention; development; turnover; Mondragon JEL classiﬁcations: M12; M53; M54

INTRODUCTION In a business environment in which people, with their knowledge, skills, experience, and motivation, are counted as a resource with great potential for generating sustainable competitive advantages, the war for talent acquires an increasingly sharper proﬁle. Among the different HR policies designed within this competitive environment to attract, develop, and retain valuable staff, training policy has gained in prominence over recent years, with a growing number of companies choosing to bolster their internal training structures. As shown in Table 1, 15% of European companies with more than 10 employees that provide continuous training have their own training centers (Cedefop, 2010, p. 37). A comparison of the 1999 and 2005 ﬁgures shows that the percentage of companies with their own training centers increased in most European countries (Table 1). This percentage rises to 34% in the case of companies with more than 250 employees. To underline the bigger strategic role given to those corporate training centers, many of them have adopted the name of corporate universities.1 This growth in the number of companies that opt to create and reinforce corporate training centers, as well as the substantial economic resources devoted to this end, contrast with the scarcity of investigations designed to

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

Table 1.

233

Businesses with More Than 10 Employees with a Training Center in 1999 and 2005. FI EL SE NL HU EE CZ DE FR LT PL RO LV IT PT ES BG LU SI

1999 (%) 14 28 16 10 2005 (%) 8 23 11 9 Difference in 6 5 5 1 % points

4 4 0

5 6 1

7 9 2

5 7 2

13 15 2

4 6 2

2 4 2

9 5 25 14 13 7 16 4 13 11 32 21 21 16 25 17 4 6 7 7 8 9 9 13

Source: CEDEFOP (2010, p. 36).

clarify the degree to which these training centers can bring advantages to their companies in terms of attracting, developing, and keeping valuable staff. The present study concentrates on the case of one of the main Spanish business groups (Mondragon Cooperative Group) that has historically stood out for its ﬁrm commitment to the creation and consolidation of corporate training centers. The phenomenon of cooperative business development led by Mondragon, unmatched at a world level, has been receiving attention by researchers from different areas since the 1970s. The central question the research set out to answer was: what lies at the root of the success of this business group? According to several authors, there are different (though not mutually exclusive) reasons that explain the overall success of the experience. The people who make up this business group, cooperative members in the main, in combination with the corporation’s training policy, are understood by certain researchers to be one of the reasons for this cooperative business success (Agirre, 2001; Aranzadi, 2003; Asua, 1988; Bradley & Gelb, 1985; Campos, 2005; Chaves, 2003; Ellerman, 1984; Hoover, 1992; Meek and Woodworth, 1990; Morris, 1992; Thomas & Logan, 1982). Other reasons for success mentioned by the literature are also linked to corporate training policy: the existence of better managers and the development of in-group management tools (Albizu & Basterretxea, 1998; Basterretxea & Albizu, 2010a; Charterina, Albizu, & Landeta, 2007; Cheney, 1999; Clamp, 2000; Jacobsen, 2001; Logan, 1988; Smith, 2001; Thomas & Logan, 1982; Whyte & Whyte, 1988) and corporate cooperative culture (Aranzadi, 2003; Bradley & Gelb, 1985; Cheney, 1999; Smith, 2001). The objective of this research work is to ascertain the degree to which corporate training centers today act as a source of perceived competitive advantage for Mondragon. We feel this study might be of interest to HR managers and heads of training in companies and entities that, like

234

I. BASTERRETXEA AND E. ALBIZU

Mondragon, face constraints in their pay policy that stand in the way of competing through wages in the ﬁght to attract and hold on to valuable staff. In addition, this investigation can throw some light on the debate around the creation of training structures by companies, constituting a new step forward in research on training policy as a source of competitive advantage. This chapter provides an update of the work previously carried out on the training structure of Mondragon. Most of the previous analyses (Asua, 1988; Bradley & Gelb, 1985; Meek & Woodworth, 1990; Whyte & Whyte, 1988) portray the educational fabric of Mondragon in the 1970s and 1980s. Therefore, they do not discuss several important strategic decisions related to the group’s training policy during the last two decades: most notably, the creation of Mondragon University (MU) and Otalora, the management training centre. Besides, this chapter differs from previous studies in the methodology used. While previous studies are mainly qualitative, based on opinions and data from a reduced number of managers of the corporate ofﬁces and some individual cooperatives, our survey offers conclusions based on a questionnaire completed by a representative sample of Mondragon HR managers. Regarding the structure of this chapter, an introduction to the research is followed by a review of works that analyze training policy as a stimulus to attract, develop, and retain valuable workers. The third part is devoted to a brief presentation of Mondragon and its training policy, followed in the fourth part by an explanation of the methodology employed in the investigation. Part ﬁve sets out the main results of the ﬁeldwork. The sixth part is given over to discussion and also looks into the contributions that this investigation makes with regard to business practice. We ﬁnish the chapter by pointing out the limitations of the study.

THE VALUE OF TALENT The staff of a company, with their knowledge, skills, experience, and motivation, are regarded as one of the most powerful resources for generating sustainable competitive advantages (Bayo & Merino, 2000, 2001; Mueller, 1996; Wright, McManaman, & McWilliams, 1994), especially in high-performance organizations (Barron, 2008). This is especially true in the managerial and technical labor markets, where growing investments in technologies and high-standard machinery modify working methods and increase the need for a skilled workforce.

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

235

Having valuable human resources, however, is no guarantee of obtaining competitive advantages unless such resources are scarce (Barney & Wright, 1998). The condition of scarcity stems from the heterogeneity of human resources. People differ in their abilities, knowledge, and attitudes and, consequently, in their contribution to the companies where they work (Wright et al., 1994). Within any labor pool, differences exist between individuals in terms of job-related skills and abilities (Barney & Wright, 1998), and it is hard to attract and keep people who guarantee highperformance levels in the organization. The war for talent is therefore a core concern for most ﬁrms. Effective strategies are nowadays being sought to achieve competitive advantages based on talent attraction and retention (Fegley, 2006). In addition, employees are adopting the philosophy that their job security lies in their employability, rather than in their employment. Talented workers with valuable skills expect employers to continually upgrade these skills, knowledge, and ability. Organizations that focus on developing talent will be in a stronger position to retain key employees (Boxall & Purcell, 2003; Capelli, 2008). Several recent investigations have dealt with the inﬂuence of training and development on a ﬁrm’s capacity to attract, retain, and develop valuable employees. Fey, Bjo¨rkman, and Pavlovskaya (2000), based on data from 101 foreign ﬁrms operating in Russia, ﬁnd that providing non-technical training is positively associated with a higher degree of development, motivation, and retention of managers. A survey conducted by Hay Group (Hay, 2002) among 330 companies in 50 countries shows that lack of training and poor career development is resulting in employees leaving and moving on. In the same sense, the results of the research done by ExecEd and MSGM’s ISL on Australian managers (Hughes, 2009) suggest that management development and training programs are effective strategies for talent retention. Jean-Marie Hiltrop (1999) analyzes HR policies and practices of 115 multinational and 204 domestic companies located in Western Europe and the impact of such policies on their capability to attract and retain employees. With the help of an independent group of management consultants, he divides those companies into three groups depending on their ability (superior, average or poor) to attract and retain talented people, and then analyzes whether those differences can be explained by a different use of HR policies. According to his study, companies with a superior ability to attract and retain talented people scored well above those in the average and

236

I. BASTERRETXEA AND E. ALBIZU

low ability categories on several HRM factors. Out of eleven different factors, the training factor ranked second in importance when explaining the difference between companies with a superior ability to attract and retain valuable employees and those with an average or poor ability. Some other studies analyze the impact of training on employee retention not in isolation, but rather by linking the training policy with other HR policies and analyzing the link between the HR system and rotation. These studies (Batt, 2002; Gelade & Ivery, 2003; Guest, Michie, Conway, & Sheenan, 2003; Huselid, 1995; Krueger & Rouse, 1998; U´beda, 2005) also point to the existence of a positive relation between training policy and lower turnover.

THE MONDRAGON CORPORATION AND TRAINING POLICY The Mondragon corporation, whose origins date back to 1956, has now become a diversiﬁed business group with 85,066 employees,2 comprising over 260 companies, half of which are cooperatives, belonging to its industrial, ﬁnance, and distribution divisions. To facilitate the corporation’s development, Mondragon is backed up by a set of organizations, consisting of a multi-level network of training centers, 12 research centers, and suprastructure entities. It had a turnover of 13,819 million euros in 2009, exporting 3,172 million euros, 59.4% of the industrial group’s production value. It is present in 15 foreign countries in the shape of 75 productive plants and has its own branches in another nine countries. It is currently the seventh largest Spanish business group. As shown in Fig. 1, Mondragon is a highly diversiﬁed corporation. The Mondragon corporation has created several corporate governing bodies (Congress, General Assembly, Governing Council, Divisions, etc.), although it should be noted that each of the cooperatives in the group is a sovereign entity. Mondragon’s corporate governing bodies and central services cannot base their power on controlling the cooperatives’ assets or shares in the cooperatives’ capital, since each cooperative’s assets and capital belong to its members. Therefore, the corporation does not own the cooperatives; rather, it is the cooperative members themselves that ultimately vote whether to join or leave the Mondragon corporation. Membership in the corporation obliges cooperatives to fulﬁll several commitments, such as the relocation of members from other cooperatives in

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

237

Presidency General Council

Areas

Vertical Transport

Finance Area

Industrial SystemsUlma

Equipment

Components

Household

Engineering and Services

Industrial Automation

Tooling and Systems

Machine tools – Danobat Group CM Automotive

Distribution Area

Automotive Chassis and Powertrain

Construction

Industry Area

Training centres

Research centres

Knowledge Area

Fig. 1.

Organizational Structure of Mondragon in 2010. Source: Mondragon (2010).

crisis; a distribution of proﬁts based on solidarity3 and reinvestment;4 the pooling of part of their returns;5 a uniformity in the initial capital contribution required from new members;6 and limitations on managers’ salaries. The salary limitations of managers in Mondragon cooperatives have always been an important source of internal controversy and external debate at Mondragon (Basterretxea, 2008; Basterretxea & Albizu, 2010a; Kasmir, 1996; MacLeod, 1997; Morris, 1992; Thomas & Logan, 1982). In 2008, only 3% of cooperative members earned a salary higher than 3.5 times that of the lower paid member (Mondragon, 2009, p.52) and, in most cooperatives, the salary of the top manager is still limited to 4.5 times more than that of the lowest paid cooperative member (Basterretxea & Albizu, 2010a). Those limited wage differentials make it difﬁcult to attract and retain managers and highly valuable professionals.

238

I. BASTERRETXEA AND E. ALBIZU

From its beginnings, the active training policy of the Mondragon group has constituted one of its most distinctive characteristics. As Arizmendiarrieta, founder of the Mondragon project, put it (Ormaetxea, 1991, p. 44), ‘‘it has been said that cooperativism is an economic movement that uses educational action, while the deﬁnition could also be altered to afﬁrm that it is an educational movement that uses economic action’’. Most analyses that set out to explain the successful evolution of the Mondragon Cooperative Experience ascribe a relevant role to the educational centers of the corporation as a group. This appraisal of corporate training centers as a source of competitive advantage or as a basic explanatory component in the Mondragon success is cited both in the analyses made by the group’s founders and successive top managers in different internal publications of the corporation, and in a good number of studies carried out by researchers from outside the Mondragon experience in the 1980s and 1990s (Asua, 1988; Basterretxea & Albizu, 2010a; Campos, 2005; Ellerman, 1984; Meek & Woodworth, 1990; Morris, 1992; Thomas & Logan, 1982; Whyte & Whyte, 1988). The Mondragon Corporation has created a complex network of educational centers that comprises a university, various vocational training centers, a business school, a cooperative and management training centre, and even children education centers at primary and secondary levels. This network is complemented by a set of associated companies and entities, which pursue objectives such as the promotion of entrepreneurship, dual training, language teaching, or company-applied research. Table 2 attempts to synthesize Mondragon’s complex educational framework. Among the various corporate training centers, a special role is played by Mondragon Unibertsitatea (MU/Mondragon University). This university, ofﬁcially set up in 1997, is supported by three educational cooperatives with a long track record: over 50 years of experience within the areas of university qualiﬁcations (in Engineering, Business and Humanities) and of Vocational Training (industrial branches). In the educational year 2007/ 2008, MU had 3,707 pupils signed up for university qualiﬁcations, 10% of whom were studying at third cycle and postgraduate level. Its offer takes in 10 high-level training studies, 24 university degree courses, 21 university postgraduate courses, and 5 third-cycle studies (PhD). In addition, there is a wide offer in continuous training, most of which is made up of short courses generated by demand from companies in the same milieu, mostly group cooperatives. The faculties that constitute MU today were home to the ﬁrst cooperative members within the Mondragon Experience and were the historical

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

Table 2.

Mondragon’s Complex Educational Framework.

Students 2007/2008

Txoriherri Lea Artibai Mondrago´n University Otalora Alecop and others

239

764 2,830 6,859 2,698 NA

Vocational Continuous University Management Training Vocational Degrees Training Training and Masters 370 362 258

394 2,403 2,894

Others

65 3,707 2,698a NA

Source: Own elaboration based on data from Mondragon. Management Training, which involved 738 people, and Co-operative Development, attended by 1,960 people.

a

suppliers of technicians, engineers, holders of diplomas and degrees in Business Administration, entrepreneurs, and managers for its cooperatives. MU students have an opportunity to combine their studies with work in the industrial cooperative Alecop, a cooperative within the Mondragon Group created in 1966, whose workforce has, since its foundation, been principally made up of students from the different centers that today form MU. Thus, more than 7,500 students have combined their studies with part-time work in this cooperative or in other Mondragon cooperatives to which Alecop assigns some of its student-workers. The bodies for management, shareholding, ownership and payment systems are practically the same in Alecop as in the remaining Group cooperatives, which means that the cooperative-based work experience of MU students at Alecop lubricates socialization in cooperative values and the future integration of technicians and university graduates in the Mondragon cooperatives (Asua, 1988; Meek & Woodworth, 1990; Thomas & Logan, 1982; Whyte & Whyte, 1988). In addition to the university, the Mondragon Group runs specialized vocational training centers (Eskola-MU-, Lea Artibai, Txori Herri) in industrial areas where local cooperatives are operative (approximately 1,000 students enrolled in formal vocational training and 5,691 people in continuous vocational training in those centers during the educational year 2007/2008). Another corporate training center that is of great importance for the cooperative group is Otalora, the Cooperative and Management Training Center of Mondragon, founded in 1984 for the purpose of training

240

I. BASTERRETXEA AND E. ALBIZU

managers and members of social bodies attached to the Mondragon cooperative group. Otalora is the training center most closely linked to the speciﬁc needs of the Mondragon cooperatives and does not offer any training to people outside the group. It offers a breadth of training designed to facilitate the internal promotion process, improve the competences of company managers, benchmarking and the dissemination of cooperative culture and values among managers and members of the control bodies. Prominent within this training offer is the MBA Executive in Cooperative Company Management. More than 500 executives and managers from Mondragon cooperatives have taken Otalora’s MBA, and practically all the current executives in the Group’s cooperatives have taken this Master course. In the educational year 2007/2008, Otalora offered Management Training involving 738 people and Cooperative Development training with the attendance of 1,960 people. The Mondragon cooperatives aid in the funding of MU, Otalora and the Group’s vocational training centers. By law, Spanish cooperatives must earmark 10% of their proﬁts to the Fund for Cooperative Education and Promotion, and Mondragon cooperatives allocate a considerable portion of those funds to the corporate training centers.7 These contributions from cooperatives are the main source of funding for Otalora, the management training center. Vocational training centers are mostly funded by Public Administration, and in the case of MU, it is the students themselves that cover most of its expenses. Besides contributing to the funding of the training centers, the group’s cooperatives play an active role in the bodies that manage and control the centers. The cooperatives that collaborate with corporate training centers have a third of the votes in the General Council and in the training centers’ Council. This allows the training content to suit the needs of the cooperatives (Basterretxea, 2008). In any case, the training offered in MU and in the corporate vocational training centers is not speciﬁc to the Mondragon cooperatives and is also valuable for work in ﬁrms outside the corporation.8 Mondragon’s commitment to training is reﬂected in the training structures that the corporation has created, as well as in the changes in the cooperative members’ qualiﬁcations. As shown in Table 3, which compares the Industrial Group’s training levels in 1987 and 2006, the change has been remarkable. In this period, the group has grown from a workforce that was mainly unskilled or had only basic qualiﬁcations to one in which most employees are at least high school graduates or have a vocational training degree.

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

Table 3.

1987 2006

241

Distribution of Mondragon Industrial Group Members According to Training (1987–2006).

Uneducated/Basic Education

High School Graduates

Vocational Training Degrees

University Graduates

63% 24.4%

3% 6.9%

23% 45.1%

11% 23.6%

Source: Own elaboration, based on Asua (1988) and Mondragon Internal Documentation (2010).

This increase in qualiﬁcations among members is mainly explained by the fact that a growing number of cooperatives have set a minimum of a vocational training degree as an admission requirement, which has allowed the percentage of vocational training graduates to double in the reference period. The percentage of members with high school qualiﬁcations and those with university degrees has more than doubled in that same period. The continuous training offered by the corporation has also contributed to that higher degree of qualiﬁcations. In the early 1990s, over 1,000 unqualiﬁed members earned vocational training degrees through the professional retraining plan implemented by Mondragon to facilitate the workforce’s multi-skill capability and the relocation of members among cooperatives in times of crisis (Basterretxea & Albizu, 2010b).

METHODOLOGY OF THE INVESTIGATION The survey was carried out with the enterprises in the group. More speciﬁcally, the population under study comprises all the cooperatives that were part of the Mondragon Group in the Basque Country and in Navarre (Spain) in December 2006, with the exception of corporate training centers and cover entities (companies created to provide services to cooperatives in the group). Also excluded were companies that have become members of the corporation over the past ﬁve years and had no previous link with the corporate training centers. Accordingly, the population was made up of a total of 81 ﬁrstdegree cooperative companies (cooperative parent companies), 66 of which collaborated with the research by replying to the survey, that is to say, 81.5% of the population.

242

I. BASTERRETXEA AND E. ALBIZU

The questionnaires, with an accompanying letter, were sent to HR managers in Mondragon, following a pre-test with various cooperative managers. In the cooperatives where this position did not exist, the questionnaire was sent to the company executive manager. Subsequently, the same standardized questionnaire acted as the basis for telephone interviews with the people who had initially been contacted by post. These telephone interviews were carried out between November 28, 2006, and January 11, 2007. The HR managers’ responses were measured using ﬁve-point subjective scales ranging from 1 (totally disagree) to 5 (totally agree). The respondents were asked to evaluate their perceptions about the link between Mondragon’s training policy and the ability of its cooperatives to attract, develop, and retain valuable employees. Many of those subjective perceptions are relative perceptual measures,9 obtained by asking the respondents to evaluate their ability to attract, develop, and retain employees (via training policy), in comparison with the ability of their competitors (e.g., ‘‘The continuous training we have provided has enabled us to develop more valuable staff than the staff of our competitors’’). Appendix A details the variables, evidence, and measures of relative perception of the questionnaire used in the chapter. Furthermore, we have also taken into account a series of classiﬁcation variables to ﬁnd out whether cooperatives that collaborate to a greater degree with Mondragon’s corporate training centers and those that put more effort into continuous training perceive greater competitive advantages (Appendix B).

RESULTS Mondragon cooperatives have a preference for collaboration with corporate training centers and particularly with MU.10 40 of the 66 cooperatives surveyed cite MU as the university they mainly collaborate with. A signiﬁcant percentage of vocational training graduates and university graduates who work in cooperatives (around 38% in both cases) come from corporate training centers, although the distribution of these centers is not uniform if we take into consideration the sectors that the companies belong to, the groups they fall into, and their location. The Mondragon cooperative collective is not homogeneous where the relation with corporate training centers and preferential recruitment from their graduates is concerned. The cooperatives that are geographically closest to the centers and those that belong to the industrial group exhibit

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

243

the highest rates of collaboration with corporate training centers and are the most inclined to resort to them when obtaining staff. The remaining cooperatives, particularly those in the retail group, scarcely collaborate with the corporate training centers, and their intake of graduates from them is lower. All told, a fourth of the cooperative managers, all of them in cooperatives in the industrial group, come from corporate training centers. The answers to the questions that measure the effort in continuous vocational training in the cooperatives (VC5 to VC8 in Appendix B) show a major commitment to training. Comparing our results with those of the Continuing Vocational Training Survey coordinated by Eurostat in 2005 and presented in 2008,11 we ﬁnd that Mondragon cooperatives put considerably more effort into the continuous training of their employees than Spanish and European ﬁrms of a similar size. In 2006, investment in continuous training in Mondragon cooperatives represented an average of 2.87% of the total wage bill.12 This 2.87% on training outlay as a percentage of labor costs more than doubles the average for Spanish companies, which is situated at 1.2%; it signiﬁcantly surpasses the 1.6% average for companies in the EU-27, and it lies among the average percentages for the companies in the European countries that are most committed to continuous training. This higher training input is also reﬂected in a higher proportion of employees who beneﬁt from training (50.5% against 33% in the EU-27) and in more hours of training per employee (23.74 hours, against the average of 12 for the EU-27). The greatest differences in training input between the Mondragon cooperatives and Spanish and European companies are found in the segment of small and medium businesses, which includes 65% of the cooperatives in our sample. Only 23% of the continuous training of cooperative workers takes place in corporate training centers. When the recipients of continuous training are cooperative managers, the training is based to a greater degree in corporate training centers. In fact, 36% of management training takes place in corporate training centers (Basterretxea, 2008, p. 497). Those differences are coherent with the different kind of training provided to workers and managers. Much of the training provided to blue collar cooperative members is generic, to foster the multi-skill capability of workers and their relocation among different cooperatives in case of crisis (Basterretxea & Albizu, 2010b). On the contrary, much of the management training, especially that provided at Otalora, is more speciﬁc to the Mondragon cooperatives. This speciﬁc training supports the retention of managers, to the extent that such training has more value in Mondragon or in other cooperatives than in other local capitalist ﬁrms (Basterretxea & Albizu, 2010a, p. 11).

244

I. BASTERRETXEA AND E. ALBIZU

Advantages for Attracting Valuable Staff To test out if Mondragon’s training policy, backed up by its corporate training centers, helps cooperatives to attract valuable staff, questions were formulated covering seven dimensions. The aggregate results of the responses given by HR managers are shown in Fig. 2. 76.6% of HR managers agree that cooperative training centers produce staff who is more familiar with cooperative culture, while 12% are indifferent and only 10.9% disagree. The data analyzed, therefore, allow us to afﬁrm that the group’s training centers fulﬁll a socializing function of cooperative values among students. These values constitute one of the relevant factors that have provided the Mondragon Experience with continuity and cohesion. In any case, it should be noted that the HR managers’ assessment regarding this question is slightly less positive among those cooperatives with a higher percentage of graduates from corporate training centers, as shown in Table 4. Likewise, Mondragon HR managers consider that candidates from corporate training centers have a more favorable attitude toward joining a cooperative enterprise as workers (53.1% in agreement and 17.2% very much in agreement). Mondragon cooperatives surpass the barriers to the attraction of valuable staff that are common to Social Economy enterprises. 46.8% of HR heads are in agreement or totally in agreement that their ability to attract valuable employees is greater than their competitors; 40.3% express no preference,

"Corporate training centres provide candidates who are more familiar with cooperative culture."

3.95

"Corporate training centres provide candidates who are more willing to work in a cooperative."

3.72

"The capacity of my company to attract valuable employees is better than that of others in the sector."

3.42

"Our relation with training centres brings competitive advantages when attracting to graduates."

3.39

“Corporate training centres provide candidates with knowledge and technical competences more suited to the needs of my company.”

3.18 2.86

"The employment services of MU and other corporate centers facilitate our recruitment and selection process."

2.48

"The employment services of MU and other corporate centers allow us to reduce the number of unsuccessful selections."

0

Fig. 2.

1

2

3

4

5

Response Frequencies for the Variables that Denote Mondragon’s Perceived Competitive Advantage in Staff Attraction.

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

245

Table 4. Correlation between Percentage of Professionals That Come from Corporate Training Centers and Advantages for the Attraction and Retention of Valuable Staff. V2 – Corporate V5 – Employment V8 – Lower Training Centers Services of MU Turnover Rates Produce Candidates Facilitate Our of Valuable More Familiar with Recruitment and Staff Than Cooperative Culture Selection Competitors VC3 – % of Pearson’s graduates in correlation Kendall’s tau-b vocational Spearman’s training from rho corporate N training centers VC4 – % of Pearson’s university correlation graduates from Kendall’s tau-b corporate Spearman’s training centers rho N

0.334**

0.261***

0.322**

0.214*** 0.284***

0.182 0.234

0.188 0.228

43

43

42

***

0.245

0.147

0.452*

0.181 0.241***

0.159 0.183

0.390 0.497*

48

47

45

*

The correlation is signiﬁcant at level 0.01 (two-sided test). The correlation is signiﬁcant at level 0.05 (two-sided test). *** The correlation is signiﬁcant at level 0.10 (two-sided test). **

which may be interpreted as a situation of competitive parity, and only 12.9% ﬁnd themselves in a situation of competitive disadvantage. When asked directly whether their relation with training centers is conducive to obtaining competitive advantages for attracting graduates, more than half of the HR managers surveyed answered afﬁrmatively. The awareness of competitive advantages deriving from the relation with training centers is signiﬁcantly greater in the collective of industrial cooperatives (3.62 of 5 against 2.1 for other cooperatives) and in those that mentioned a corporate training centre as the centre they prefer to collaborate with (3.65 against 2.96 for cooperatives that collaborate with other training centers).13 The advantages deriving from technical training that is more suited to the needs of companies are not so visible for the cooperatives surveyed taken as a whole, although they are perceived by a signiﬁcant percentage of HR heads. 38.7% are of the opinion that their corporate training centers produce candidates with competences and knowledge that are more in tune

246

I. BASTERRETXEA AND E. ALBIZU

with their needs, against 17.7% who consider the opposite and 43.5% who express no preference. The assessment is better in the case of cooperatives that mentioned MU as the university they collaborate with ﬁrst and foremost (3.34 of 5, against 2.92 of cooperatives that collaborate with other universities).14 When asked whether the employment services of corporate training centers facilitate the cooperatives’ recruitment and selection processes, opinion is divided among the companies surveyed (46% unfavorable, including 19% absolutely unfavorable, against 42% completely favorable). The assessment is signiﬁcantly better in cooperatives that collaborate ﬁrst and foremost with corporate training centers (3.28 of 5, against 2.17 for the cooperatives that collaborate with other training centers), in industrial cooperatives (3.04 of 5 against 1.56 for other cooperatives)15 and in cooperatives with a higher percentage of graduates in vocational training from corporate training centers (Table 4). When asked whether the corporate training center’s employment services contribute to a reduction in the number of unsuccessful staff selections, the response is negative for 57% of the HR managers surveyed. Despite this globally negative assessment, the evaluation is again signiﬁcantly better in cooperatives that collaborate ﬁrst and foremost with corporate training centers16 and in those that belong to the Industrial Group.17

Advantages for Retaining Valuable Staff Fig. 3 shows the average values of the responses Mondragon HR managers gave to questions designed to evaluate whether the group’s training policy and corporate training centers help cooperatives to retain valuable staff. Almost 70% of Mondragon cooperatives can rely on competitive advantages for the retention of valuable staff when compared with companies in the same competitive environment, while another 17.7% ﬁnd themselves in a situation of competitive parity. This advantage vis-a`-vis the retention of valuable staff that can be appreciated in this work conﬁrms that Mondragon has been able to overcome the barriers hindering staff retention that have been pointed out in Social Economy literature. Regarding the degree to which training policy based on corporate training centers helps in the securing of this competitive advantage for retaining valuable staff, we ﬁnd conﬂicting evidence. Lower rates of turnover of valuable staff are correlated with various indicators for training policy based on corporate training centers (Table 4).

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

"The turnover rates of valuable staff are lower than in companies within the same competitive environment.”

3.74

"Offering more continuous training than our competitors helps us to retain valuable staff."

3.56

"The turnover of professionals from corporate training centers is lower than that of those trained in centrers outside of the Mondragon network."

3.06 0

Fig. 3.

247

1

2

3

4

5

Frequencies of Response to the Variables that Denote Perceived Competitive Advantage of Mondragon in Staff Retention.

Thus, cooperatives with a higher percentage of professionals from corporate vocational training centers exhibit lower rates of relative turnover than the rest of the cooperatives. This correlation is positive and signiﬁcant at 5%. The ability to hold on to valuable staff is also greater the higher the proportion of graduates there are from MU. This correlation is positive and signiﬁcant at 1%. There are also differences between the cooperatives that have mentioned MU as the university they mostly work with, and all the rest. In the former, the perception of advantages for retaining valuable staff is greater. Similarly, 62.5% of Mondragon HR managers consider that their cooperative provides more continuous training than their competitors and that this commitment to training facilitates the retention of valuable staff. The perception of competitive advantages for retaining valuable staff is also linked to different indicators of effort put into continuous training by cooperatives. The greater the percentage of employees who have received continuous training over the previous year, and the more hours are devoted to training per employee, the greater the perception of HR managers that they possess advantages for retaining their professionals. These perceived advantages, though, are not correlated with economic investment on continuous training, measured as a percentage of the wage bill that such investment represents (Table 5). The consideration of continuous training as a factor that reduces the turnover of valuable staff in cooperatives is not determined by such training being given in corporate training centers. In fact, we have not detected a signiﬁcant correlation between the perception of advantages for retaining valuable staff and the percentage of continuous training provided in collaboration with Mondragon’s corporate training centers.

0.342* 0.406* 59 0.0850 0.0502 0.0609 53 0.073 0.111 0.142 59

0.188*** 0.232*** 62 0.388* 0.318* 0.404* 53 0.170 0.143 0.181 60

**

The correlation is signiﬁcant at level 0.01 (two-sided test). The correlation is signiﬁcant at level 0.05 (two-sided test). *** The correlation is signiﬁcant at level 0.10 (two-sided test).

*

0.107 0.133 41 0.408*

0.197

0.050 0.068 39 0.239***

0.070

V8 – Lower V12 – Turnover Rates Continuous of Valuable Training Staff Than Enabled More Competitors Valuable Staff Than Competitors

0.0853 0.1089 53 0.009 0.100 0.136 62

0.213** 0.269** 60

0.468* 0.566* 63 0.0691

0.066 0.073 42 0.569*

0.048

V15 – Continuous Training Produced Greater Speed of Response to Customers

0.0529 0.0763 51 0.223***

0.318* 0.388* 62 0.1616

0.020 0.023 41 0.385*

0.003

V14 – Continuous Training Has Brought a Reduction in Costs

0.152 0.199 60

0.1150 0.1397 51 0.143

0.419* 0.510* 60 0.0671

0.187 0.230 40 0.508*

0.179

0.191 0.238 62

0.1536 0.1883 53 0.165

0.379* 0.466* 63 0.1227

0.077 0.092 42 0.469*

0.088

V16 – V17 – Continuous Continuous Training Brought Training Improvement in Produced Quality Reduction in Customer Complaints

Correlation between Indicators of Effort on Continuous Training and Advantages in the Development of Valuable Staff.

VC5 – % training Pearson’s expense/wage costs correlation Kendall’s tau-b Spearman’s rho N VC6 – rate of Pearson’s participation of correlation Kendall’s tau-b employees in continuous Spearman’s rho N training VC7 – training Pearson’s hours/employee correlation Kendall’s tau-b Spearman’s rho N VC8 – % continuous Pearson’s training provided correlation Kendall’s tau-b with corporate Spearman’s Rho training centers N

Table 5.

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

249

In contradiction with the results that indicate a positive relation between training policy based on corporate training centers and securing advantages for the retention of valuable staff, less than a third (32%) of HR managers surveyed considered that professionals from corporate training centers are more loyal to cooperatives, while the majority expressed no preference with regard to this statement (43.4%). It must be stressed that there are no signiﬁcant differences in this assessment in terms of whether the cooperative collaborates to a greater or lesser degree with the Mondragon training centers, or whether it has a greater or lesser intake of their graduates.

Developing Valuable Staff The averages of the replies to the questions asked to assess whether Mondragon’s training policy and training centers help cooperatives to develop valuable staff are reﬂected in Fig. 4. Continuous training has had a positive effect on the motivation of workers, according to 74.6% of Mondragon heads of HR. For a vast majority (76.2%) of Mondragon HR managers, continuous training has led to better quality in the products and services of their companies. Most of their cooperatives have adopted quality strategies, with a great deal of effort and notable success. What emerges from the results of this survey is that the cooperatives’ training policy has been in alignment with the quality strategy, and has had a positive inﬂuence on it. The "Continuous training has provided our workers with greater motivation.”

3.81

"Continuous training has brought about an improvement in the quality of our products and services.”

3.79

"Having corporate training centers enables us to provide continuous training that is better suited to our needs.”

3.77

"The continuous training we have privided has enabled us to develop more valuable staff than the staff of our competitors.”

3.56

"Continuous training has produced a greater speed of response to customers.”

3.35

"Continuous training has produced a reduction in customer complaints.”

3.25 3.2

"Continuous training has brought a reduction in costs.”

0

Fig. 4.

1

2

3

4

5

Frequencies of Response to the Variables that Denote Perceived Competitive Advantage for Mondragon in Staff Development.

250

I. BASTERRETXEA AND E. ALBIZU

cooperatives that provide training for a greater proportion of employees are the ones that more positively evaluate the effect of training on quality (Table 5). This highly signiﬁcant positive correlation is because the management model and the total quality systems implemented by the Mondragon cooperatives require the active participation of the greatest possible number of employees in continuous improvement. When the managers were asked whether Mondragon’s corporate training centers are conducive to continuous training that is more suited to the needs of the group’s companies, more than two-thirds (67.7%) of the heads of HR at Mondragon were in agreement that this is so. This positive assessment of corporate training centers does not differ in terms of the degree to which the cooperatives collaborate with them in order to provide their employees and their managers with continuous training. As well as producing more motivated staff, more than half of those surveyed (54.2%) consider that continuous training has enabled their cooperatives to develop more valuable staff than their competitors. This perceived competitive advantage increases the higher the percentage of employees take part in continuous training actions (Table 5). A signiﬁcant percentage, though not the majority, of the HR managers consulted consider continuous training to have had a positive effect on greater customer satisfaction. Speciﬁcally, 44.5% of those surveyed state that training has facilitated a more rapid response to customer requirements, and 39.3% that training has brought about a reduction in the number of such complaints. Once again, the factor that best explains why continuous training leads to better customer service is the fact that training is offered to a greater proportion of employees (Table 5). For 37.7% of the managers interviewed, continuous training has contributed to a reduction of costs. This cost decrease due to permanent training is particularly perceived in cooperatives that extend their continuous training program to a greater percentage of their employees, and in those that rely more on corporate training centers to cover their continuous training needs.

DISCUSSION The Mondragon cooperatives overcome the barriers against attracting valuable staff that are common to Social Economy enterprises and, in the main, evaluate their situation as one of parity or of competitive advantage in relation to companies in their sectors in terms of being able to attract valuable employees.

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

251

Mondragon’s training policy, backed up by its corporate training centers, is internally seen as a source of competitive advantages for attracting valuable staff, fundamentally due to a supply of candidates who are more socialized in cooperative culture and more inclined to want to work in cooperatives. These perceived advantages generated by corporate training centers are conﬁned to the cooperatives in the Industrial Group. Mondragon’s HR managers consider that continuous training policy produces more motivated and more valuable staff than is the case for competitors. They also consider that continuous training has led to improvements in quality and, to a lesser extent, a reduction in costs and better customer service. The perceived competitive advantages deriving from continuous training in the Mondragon cooperatives are positively correlated with the percentage of employees receiving such training. However, we have not found a correlation with other indicators of effort on continuous training, such as the number of hours of training per employee, or training budget in relation to wage costs. The Mondragon cooperatives evaluate their situation, in the main, as one of competitive advantage with regard to companies in their ﬁeld when it comes to holding on to valuable staff. Such perceived competitive advantage is to a large extent explained by the formula of offering more continuous training provision than their competitors. It is not clear, however, that the training background of workers (cooperative vs. non-cooperative training centers) is directly related to the propensity of workers to move on to other employment. The link between the provision of training and retention can be explained, in the case of training provided to managers, by its high speciﬁcity. This speciﬁc training supports the retention of managers, to the extent that such training has more value in Mondragon cooperatives than in investor-owned ﬁrms. In the case of the training provided to blue collar cooperative members, training is more general than in investor-owned ﬁrms. This general training fosters the multi-skill capability of the workforce and makes it easier to relocate members in different jobs and cooperatives in times of crisis (Basterretxea & Albizu, 2010a, 2010b). Thus, general training provided to workers enhances job security. Since job security is the most valued factor by cooperative members in the employee satisfaction surveys carried out in Mondragon (Basterretxea, 2008), we can suggest that general training enhances employee retention when it generates a higher degree of employment security. The link between training and retention can also be explained by the internal promotion that follows many training processes. This is especially

252

I. BASTERRETXEA AND E. ALBIZU

true in the case of training provided to managers, since management training and promotion policies are regarded as inseparable policies in Mondragon. Another factor that can explain the retention of employees after training is the existence of exit barriers. Anyone seeking membership in a Mondragon cooperative is aware of the risk of losing part of the initial contribution to the cooperative’s share capital and the subsequent accumulated dividends if they leave the cooperative. A fraction of their initial capital contribution (up to 20%) goes straight into the cooperative’s reserve fund and is not refundable on an individual basis in any case. Besides, when members leave, the cooperative can deduct considerable sums from their initial contribution and from their account’s accumulated dividends, a deduction that will vary according to how much the cooperative invested in the member’s training (Gorron˜ogoitia, 1991; Larran˜aga, 1991). The time factor and asset mass efﬁciency constitute barriers against any imitation of the Mondragon corporate training centers. The combined effect of time compression diseconomies, experience economies and time ﬂow trajectories make the time factor an important obstacle for any competitor wishing to create a corporate training centre structure or mechanisms for company-centre cooperation as efﬁcient as those created by Mondragon over decades. A training policy based on corporate training centers, which provides Mondragon cooperative companies with so many perceived advantages, may also be a source of competitive disadvantages in the future because of the speciﬁcity of resources involved. The Mondragon network of corporate training centers (and the degrees and diplomas they provide) has its roots to a large extent in the historical needs of industrial cooperatives in the ﬁeld of human resources. This speciﬁcity means that some cooperatives that are in a process of expansion, such as those in the Retail Group, have historically beneﬁted little from the corporate training centers. Likewise, this speciﬁcity acts as a hindrance to the corporate training centers being able to provide assistance in launching and driving some of the new business lines that Mondragon aims to give impetus to in its present strategic reorientation. This calls into question some aspects of the training policy that has been pursued until now and suggests that the corporation should reconsider the keystones of its relation with the corporate training centers. This chapter contributes to the literature on the educational fabric of Mondragon adding updated empirical evidence to previous studies and incorporating the point of view of HR managers of the group’s cooperatives. Our results show that Mondragon HR managers consider that the corporate training policy and training centers generate many advantages, but we have

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

253

also found such advantages to be more moderate than those cited in the historical literature on Mondragon. The broader source of data and opinions of the present study shows a less homogeneous link between corporate training centers and cooperatives than in previous studies. Likewise, we have found that some sources of advantage according to previous literature (technical training more suited to the needs of cooperatives, training centers as facilitators of recruitment, lower rotation of graduates in corporate training centers) are not perceived by the cooperatives’ HR managers. With respect to the contribution of this chapter to the literature on training policy, the chapter’s ﬁndings show that training policy and corporate training centers can be conceived by HR managers as a source of advantages for the attraction, development, and retention of staff. Most studies on training policy analyze the impact of training on business results, but there are few studies that explore the link between training and employee attraction and retention. The chapter’s ﬁndings, in particular those regarding the effect of training on worker attraction and retention, add empirical evidence to the few studies on the subject. Besides, our results suggest that training policy can have a positive impact on employee attraction and retention in a context of serious limitations and particularities in other HR policies, such as wage policy. Other large business corporations that are presently thinking of making a commitment to strengthening their training structures or creating corporate training centers on the same lines as those analyzed in this work would have to take into account that such a move only makes sense in interconnection with other HR policies. In the experience that has been analyzed here, high investment in training takes on more meaning when internal job markets and extensive job ladders exist in each cooperative and in the corporation. Moreover, this is a corporation in which business culture is the main point of union between the cooperatives that form it, which demands extra input into corporate culture training. Another feature of this corporation’s HR policy is the existence of wage limitations, and such a restriction stands in the way of using pay as a preferential means of attracting and retaining valuable staff. In companies where business culture is not so important, where the possibilities of promotion are relatively slim, and where it is easy to attract and retain talent with competitive wages, it probably does not make sense to set up a corporate university and other training centers similar to those analyzed in this study. Finally, any possible imitator would have to be aware of the high cost that could be incurred by the creation of a training centre structure similar to

254

I. BASTERRETXEA AND E. ALBIZU

that of Mondragon. In the case we have studied here, this cost is, to a large extent, defrayed as a result of the favorable tax treatment that is accorded in Spain to investments by cooperatives on continuous training and on their corporate training centers. Consequently, even if a competitor were to develop continuous training programs and corporate training centers as valuable as those of Mondragon, the costs the competitor would have to incur would cancel out part or all of that value.

LIMITATIONS The main limitations of the study stem from the subjectivity of the data used (opinions of HR managers about the effectiveness of training policies). This subjectivity bias can be especially important in those questions in which HR managers evaluate training policies that they have themselves supported or implemented (continuous training policies in the main). In order to reduce this subjectivity, it would be interesting to conduct further research to learn what kind of answers to the same questions would be provided by workers. Another limitation is that we have not considered possible reverse causality relationships. In some questions in which most HR managers strongly disagreed with some factors as a source of advantages (e.g., when asked whether MU employment services facilitate their recruitment process), we have highlighted that those cooperatives with a higher degree of collaboration with corporate training centers had a more positive point of view. In those questions, there could also be reverse causality: the negative assessment of most of HR managers on some features of Mondragon training centers (e.g., their employment services) may explain a lower degree of collaboration with such centers.

NOTES 1. In the United States, there is a growing number of corporate universities in virtually all sectors. American Express, Apple, Bell Atlantic, Disney, Ford, General Motors, Intel, Harley Davidson, McDonalds, and Sun Microsystems are just some of these companies with corporate universities. In recent years, major European ﬁrms such as AXA, Bayard Press, Bouygues, Cap Gemini, Ernst & Young, Daimler, Heineken, Lufthansa, Siemens and Schneider Electric have also strengthened their corporate training centers by adopting the corporate university name. 2. 38.4% in the Basque Country, 44.5% in the rest of Spain and 17% in the rest of the world (Mondragon, 2010).

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

255

3. 10% of proﬁts are allocated to the Fund for Cooperative Education and Promotion. This fund promotes the training of members and workers, intercooperative relations, and cultural and social welfare programs. Cooperatives earmark 20% of this fund for the Fund for Inter-Cooperative Education and Promotion, which gives ﬁnancial support to the corporate training centers. 4. 45% of proﬁts are allocated to the cooperative’s Reserve Fund. The remaining 45% is capitalized into each member’s account. 5. Cooperatives must pool between 15% and 40% of their returns alongside the rest of cooperatives belonging to the same division (see the specialization areas in Fig. 1). In addition, they must earmark 10% of their returns for the Mondragon Investment Fund. 6. Historically, new members of Mondragon cooperatives have made an initial contribution to the share capital of no less than the equivalent to the annual salary of the lowest paid member. In 2008, this amounted to 14,000 euros. 7. For the allocation of funds in different years, see Basterretxea (2008). 8. According to the Basque Employment Service (Lanbide, 2004), nearly 40% of the engineers from Mondragon University entered a cooperative in 2000 and so did 27% of those who had completed Management studies. This means that most of the graduates entered a company that did not belong to the group. 9. While it is true that perceptual data may introduce limitations through increased measurement error and the potential for mono-method bias, the beneﬁts outweigh the drawbacks (Fey et al., 2000). Furthermore, there is precedent for using such perceptual measures in similar research (e.g., Delaney & Huselid, 1996; Fey et al., 2000; Youndt, Snell, & Dean, 1996). Additionally, prior research has shown that subjective measures of ﬁrm performance are well correlated with objective measures (Geringer & Hebert, 1991; Powell, 1992). 10. See answers to the questions VC1 to VC4 in Appendix B. 11. See http://epp.eurostat.ec.europa.eu 12. If we consider Mondragon Corporation as a whole, and not only the cooperatives of the group, the percentage of training expenditure over total personnel costs was much lower this year (0.6%). This difference may be due to sample bias in our survey (only 40 HR managers of the cooperatives were able to quantify the percentage in the survey, while virtually everyone knew the percentage of employees who had been trained, or the number of training hours per employee). This divergence of ﬁgures may also be explained by a high concentration of training expenditure in the cooperatives of the group, while spending on training is much smaller in subsidiaries. 13. Differences signiﬁcant to 5% (t-test and Chi-square test). 14. Difference signiﬁcant to 5% (Chi-square test). 15. Differences signiﬁcant to 1% (t-test and Chi-square test). 16. Difference signiﬁcant to 1% (t-test and Chi-square test). 17. Difference signiﬁcant to 5% (t-test).

ACKNOWLEDGMENTS We thank the referees and the editor for their helpful and inspiring comments.

256

I. BASTERRETXEA AND E. ALBIZU

REFERENCES Agirre, A. (2001). Los principios cooperativos ‘‘atractores’’ de la gestio´n eﬁciente: Su medicio´n. Aplicacio´n al caso de Mondrago´n Corporacio´n Cooperativa. CIRIEC-Espan˜a, Revista de Economı´a Pu´blica, Social y Cooperativa, 39(11), 93–113. Albizu, E., & Basterretxea, I. (1998). Flexibilidad laboral y generacio´n de empleo en tiempos de crisis. El caso de Mondrago´n Corporacio´n Cooperativa. Revista Europea de Direccio´n y Economı´a de la Empresa,, 7(3), 83–98. Aranzadi, D. (2003). El signiﬁcado de la Experiencia Cooperativa de Mondrago´n. Estudios Empresariales, 112(2), 58–69. Asua, B. (1988). Educacio´n y trabajo en la sociedad industrial del Paı´s Vasco. La ‘‘Eskola Polite´cnica Jose´ Marı´a Arizmendiarreta’’ en el Grupo Cooperativo Mondrago´n. Unpublished PhD Thesis. University of the Basque Country, Leioa. Barney, J. B., & Wright, P. M. (1998). On becoming a strategic partner. The role of human resources in gaining competitive advantage. Human Resource Management, 37(1), 31–46. Barron, P. (2008). Education and talent management: Implications for the hospitality industry. International Journal of Contemporary Hospitality Management, 20(7), 730–742. Basterretxea, I. (2008). La polı´tica de formacio´n como fuente de ventaja competitiva en la experiencia Mondrago´n. Un ana´lisis desde la visio´n basada en los recursos. Unpublished PhD Thesis. University of the Basque Country, Bilbao. Basterretxea, I., & Albizu, E. (2010a). Management training as a source of perceived competitive advantage: The Mondragon Cooperative group case. Economic and Industrial Democracy, 25(October). 10.1177/0143831 10377809. Basterretxea, I., & Albizu, E. (2010b). ¿Es posible resistir a la crisis? Un ana´lisis desde la gestio´n de las polı´ ticas de empleo y formacio´n en Mondrago´n. CIRIEC-Espan˜a, Revista de Economı´a Pu´blica, Social y Cooperativa (67), 75–96. Batt, R. (2002). Managing customer services: Human resource practices, quit rates, and sales growth. Academy of Management Journal, 45(3), 587–597. Bayo, A., & Merino, J. (2000). Human resource management and operational performance in Spanish manufacturing industry. Working Paper No. 43/00. Public University of Navarra, Pamplona. Bayo, A., & Merino, J. (2001). Size and HRM in the Spanish manufacturing industry. Employee Relations, 23(2), 188–206. Boxall, P., & Purcell, J. (2003). Strategy and human resource management. Palgrave Macmillan: Basingstoke. Bradley, K., & Gelb, A. (1985). Cooperation at work: The Mondragon experience. Heinemann: London. Campos, R. S. (2005). Mondragon Unibertsitatea: entre Universidade e Formac- ao Proﬁssional. Revista Uruta´gu¨a-Revista acade´mica multidisciplinar, (6), Universidade Estadual de Maringa´. Retrieved from http://www.uem.br/urutagua/006/06campos.htm. Capelli, P. (2008). Talent management for the twenty-ﬁrst century. Harvard Business Review, 86(3), 74–81. CEDEFOP. (2010). Employer-provided vocational training in Europe: Evaluation and interpretation of the third continuing vocational training survey. Luxembourg: Publications Ofﬁce of the European Union. Charterina, J., Albizu, E., & Landeta, J. (2007). The quality of management in Basque companies: Differences existing between cooperative and non-cooperative companies.

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

257

In: S. Novkovic & V. Sena (Eds.), Advances in economic analysis of participatory and labor-managed ﬁrms. Cooperative Firms in Global Markets: Incidence, Viability and Economic Performance. (Vol. 10, pp. 109–150). New York: Elsevier Science. Chaves, R. (2003). El modelo cooperativo de Mondrago´n. El desarrollo econo´mico basado en el cooperativismo. In: R. Chaves, J. L. Monzo´n & A. Sajardo (Eds.), Elementos de economı´a social. Teorı´a y realidad (pp. 93–106). Valencia: University of Valencia. Cheney, G. (1999). Values at work: Employee participation meets market pressure at Mondrago´n. Ithaca, NY: ILR Press. Clamp, C. A. (2000). The internationalization of Mondragon. Annals of Public and Cooperative Economics, 71(4), 557–577. Delaney, J. T., & Huselid, M. A. (1996). The impact of human resource management practices on perceptions of organizational performance. Academy of Management Journal, 39(4), 949–969. Ellerman, D. P. (1984). Entrepreneurship in the Mondragon cooperatives. Review of Social Economy (42), 272–294. Fegley, S. (2006). 2006 talent management survey report, SHRM Research Quarterly. Alexandria, VA. Fey, C. F., Bjo¨rkman, I., & Pavlovskaya, A. (2000). The effect of human resource management practices on ﬁrm performance in Russia. International Journal of Human Resource Management, 11(1), 1–18. Gelade, G. A., & Ivery, M. (2003). The impact of human resource management and work climate on organizational performance. Personnel Psychology, 56(2), 383–404. Geringer, M. J., & Hebert, L. (1991). Measuring performance of international joint ventures. Journal of International Business Studies (28), 249–263. Gorron˜ogoitia, A. (1991). La organizacio´n interna de una cooperativa (Textos ba´sicos de Otalora. Chapter III). Mondrago´n: Otalora. Guest, D., Michie, J., Conway, N., & Sheenan, M. (2003). Human resource management and corporate performance in the UK. British Journal of Management, 41(2), 291–314. Hay, M. (2002). Strategies for survival in the war of talent. Career Development International, 17, 52–55. Hiltrop, J. M. (1999). The quest for the best: Human resource practices to attract and retain talent. European Management Journal, 17(4), 422–440. Hoover, K. R. (1992). Mondragon’s answers to utopia’s problems. Utopian Studies, 3(2), 1–19. Hughes, B. (2009). What will it take to keep our best people. Sydney: Macquarie Graduate School of Managements Executive Education. Huselid, M. A. (1995). The impact of human resource management practices on turnover, productivity and corporate ﬁnancial performance. Academy of Management Journal, 8(3), 635–672. Jacobsen, G. (2001). Cooperative and training dimensions in entrepreneurship. A study of the methodology of the Saiolan Centre in Mondragon. Inussuk. Artic Research Journal (1), 137–146Retrieved from http://http://cog.kent.edu/lib/JakobsenCoop.htm. Kasmir, S. (1996). The myth of Mondragon: Cooperatives, politics, and working-class life in a Basque town. New York, NY: State University of New York Press. Krueger, A., & Rouse, C. (1998). The effect of workplace education on earnings, turnover and job performance. Journal of Labor Economics, 16(1), 61–94.

258

I. BASTERRETXEA AND E. ALBIZU

Lanbide. (2004). Mondragon Unibertsitatea. Encuesta de insercio´n laboral de titulados y tituladas. Promocio´n 2000. Vitoria-Gasteiz: Lanbide. Larran˜aga, J. M. (1991). El socio cooperativista ((Chapter IV)) [Textos ba´sicos de Otalora]. Mondrago´n: Otalora. Logan, C. G. (1988). An experiment that continues: The Mondragon co-operatives. In: Congreso sobre el cooperativismo y la economı´a social en el mundo. II Congreso Mundial Vasco (pp. 111–113). Vitoria-Gasteiz: Publishing Service of the Basque Government. MacLeod, G. (1997). From Mondragon to America: Experiments in community economic development. Sidney: University College of Cape Breton Press. Meek, C. B., & Woodworth, W. P. (1990). Technical training and enterprise: Mondragon’s educational system and its implications for other cooperatives. Economic and Industrial Democracy (11), 508–528. Mondragon. (2009). 2008 annual report. Mondrago´n: Mondrago´n. Mondragon. (2010). 2009 annual report. Mondrago´n: Mondrago´n. Morris, D. (1992). The Mondragon system: Cooperation at work, institute for local self-reliance, Washington, DC. Retrieved from http://www.newrules.org/resources/MondragonCo-op.pdf Mueller, F. (1996). Human resources as strategic assets: An evolutionary resource-based theory. Journal of Management Studies, 33(6), 757–785. Ormaetxea, F. (1991). Eskola Politeknikoa Jose Maria Arizmendiarreta. Sdad. Coop.. Aretxabaleta: Otalora. Powell, T. C. (1992). Organizational alignment as a competitive advantage. Strategic Management Journal (13), 119–134. Smith, S. C. (2001). Blooming together or wilting alone? Network Externalities and the Mondragon and La Lega cooperative networks. Helsinki: United Nations University/ WIDER Discussion Paper No. 2001/27. Thomas, H., & Logan, C. (1982). Mondragon, an economic analysis. Winchester, MA: Allen & Unwin. U´beda, M. (2005). Training and business performance: The Spanish case. International Journal of Human Resource Management, 16(9), 1691–1710. Whyte, W. F., & Whyte, K. K. (1988). Making Mondrago´n. Ithaca, NY: ILR Press. Wright, P. M., McManaman, G. C., & McWilliams, A. (1994). Human resources and sustained competitive advantage: A resource-based perspective. International Journal of Human Resource Management, 5(2), 301–326. Youndt, M. A., Snell, S. A., & Dean, J. W. (1996). Human resource management, manufacturing strategy, and ﬁrm performance. Academy of Management Journal, 39(4), 836–866.

Does Training Policy Help to Attract, Retain, and Develop Human Resources?

259

APPENDIX A The variables included in the questionnaire are the following: V1 – ‘‘The capacity of my company to attract valuable employees is better than that of others in the sector.’’ V2 – ‘‘Corporate training centers produce candidates who are more familiar with cooperative culture.’’ V3 – ‘‘Corporate training centers produce candidates who are more willing to work in a cooperative.’’ V4 – ‘‘Corporate training centers produce candidates with knowledge and technical competences more suited to the needs of my company.’’ V5 – ‘‘The employment services of MU and other corporate centers facilitate our recruitment and selection process.’’ V6 – ‘‘The employment services of MU and other corporate centers allow us to reduce the number of unsuccessful selections.’’ V7 – ‘‘Our relation with training centers provides us with competitive advantages when attracting graduates.’’ V8 – ‘‘The turnover rates of valuable staff are lower than in companies within the same competitive environment.’’ V9 – ‘‘The turnover of professionals from corporate training centers is lower than that of those trained in centers outside of the Mondragon network.’’ V10 – ‘‘Offering more continuous training than competitors helps us to retain valuable staff.’’ To measure whether continuous training produces more valuable staff, in addition to the perceptions of HR managers, we included a series of indirect indicators that are usually applied in scientiﬁc literature evaluating the efﬁciency of company training activities. These indicators indirectly measure whether training produces more valuable staff, through the consumer’s value perceptions. Thus, continuous training produces more valuable staff, if the latter are able to produce better quality results, providing better customer service, or contributing to a reduction in costs. Therefore, the variables included to test this question are the following: V11 – ‘‘Having corporate training centers enables us to provide continuous training that is better suited to our needs.’’ V12 – ‘‘The continuous training we have provided has enabled us to develop more valuable staff than the staff of our competitors.’’ V13 – ‘‘Continuous training has provided our workers with greater motivation.’’

260

I. BASTERRETXEA AND E. ALBIZU

V14 – ‘‘Continuous training has brought a reduction in costs.’’ V15 – ‘‘Continuous training has produced a greater speed of response to customers.’’ V16 – ‘‘Continuous training has produced a reduction in customer complaints.’’ V17 – ‘‘Continuous training has brought about an improvement in the quality of our products and services.’’

APPENDIX B Classiﬁcation variables: VC1 – Priority collaboration with corporate vocational training centers. VC2 – Priority collaboration with Mondragon Unibertsitatea. VC3 – % of graduates in vocational training from corporate training centers. VC4 – % of university graduates from corporate training centers. VC5 – % training expense/wage costs. VC6 – Rate of participation of employees in continuous training. VC7 – No. of hours of training/employee. VC8 – % continuous training provided in corporate training centers.

GOVERNANCE AND BEHAVIOR IN NONPROFITS: ANALYSIS OF URUGUAYAN HEALTH CARE ORGANIZATIONS Juan Jose Barrios and Mieke Meurs ABSTRACT Literature on nontraditional ﬁrms has focused on behavioral differences with for-proﬁt ﬁrms. Less attention has been given to the variations in behavior among nontraditional ﬁrms. This chapter examines differences across three types of Uruguayan nonproﬁt health care organizations. This chapter draws on a unique dataset of Uruguayan health care organizations during the period 1982–1990, as well as interviews with doctors working in the three types of nonproﬁts during spring 2010. We use a simple OLS regression to identify differences in average behavior, and differences in reaction to a regulatory change. The chapter shows that structure of stake holding and governance signiﬁcantly affect behavior, even where many behaviors are highly regulated. These ﬁndings highlight the importance of specifying governance structure when predicting behavior of nontraditional ﬁrms. Empirical tests of behavioral differences between traditional and nontraditional Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 261–286 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012014

261

262

JUAN JOSE BARRIOS AND MIEKE MEURS

ﬁrms will be more meaningful if the governance structure of nontraditional ﬁrms is common and speciﬁed. A limitation of our study is our inability to control for the timing of degeneration of producer cooperatives. This would be one element of governance structure to consider in future data collection. These ﬁndings highlight the need to avoid drawing broad policy conclusions from the behavior of a speciﬁc subset of nontraditional ﬁrms. This chapter highlights the importance of carefully specifying stakeholder and governance structure when predicting behavior of nontraditional ﬁrms. It is of interest to anyone using a sample of nontraditional ﬁrms to test general hypotheses about their behavior.

INTRODUCTION Much of the literature on the behavior of nontraditional ﬁrms (nonproﬁts and cooperatives) has focused on predicted differences with traditional, forproﬁt ﬁrms. Less attention has been paid to variations in behavior among different types of nontraditional ﬁrms, resulting from the structure of stake holding or ways that stakeholders participate in governance. This omission may lead to unfounded conclusions about behavioral or performance differences between traditional and nontraditional ﬁrms. For example, in a 1992 study of the impact of property rights on efﬁciency, Fizel and Nunnikhoven use cross-sectional data to compare for-proﬁt and nonproﬁt nursing homes without specifying nonproﬁt type. They conclude that for-proﬁts are more efﬁcient than nonproﬁts in general – relative performance is not linked to a particular form (or forms) of nonproﬁt. In a 1993 study, L.L. Peters performs a meta-analysis of performance of for- and nonproﬁt electric companies. Although he notes the importance of possible variation in performance by type of nonproﬁt, the nature of the previous work forces him to consider municipal and cooperative ﬁrms together as ‘‘nonproﬁts.’’ Peters ﬁnds no signiﬁcant difference in performance between for- and nonproﬁt ﬁrms. In recent years, more attention has begun to be paid to the question of how nonproﬁt form affects behavior. Avner Ben-Ner and Benedetto Gui (2003) provide a careful typology of organization type, focusing on variation by residual claimancy and control, and discuss behavioral implications of the differences. Mehdi and Filippini (2004) empirically

Governance and Behavior in Nonproﬁts

263

examine differences in performance between publically and privately owned nonproﬁt nursing homes in Switzerland. In this chapter, we extend this literature. We examine the behavior of three types of Uruguayan private, nonproﬁt health care organizations and test for behavioral differences across the three types. We examine both differences in overall patterns of spending by the organizations and differences in the response of the organizations to an exogenous shock – a brief liberalization of government controls on membership fees. Uruguayan providers of primary health care services (outside the public sector which covers mainly the indigent, military families, and other dependents of the state) are required by a 1981 law to organize as nonproﬁts. They may choose among three types of organization: mutual beneﬁt organizations (similar to consumer cooperatives), and two types of producer (doctors) cooperatives. Although they differ in their stakeholder and governance structures, all three types of organization face a very tight regulatory environment. The state sets prices for organization membership and other fees and mandates the range of services offered. Salaries, working conditions, and hours of doctors and technical staff are centrally negotiated with unions of medical workers. Because nonproﬁts do not compete in the market for basic health care services with for-proﬁt providers, this context provides a view of ‘‘pure’’ nonproﬁt behavior, not inﬂuenced by competition with for-proﬁts. At the same time, the strict regulatory environment should reduce variations in behavior across nonproﬁts. If signiﬁcant differences in behavior exist in this regulatory context, more signiﬁcant differences might be expected between types of nonproﬁts in more liberalized contexts. We draw on a unique dataset of Uruguayan health care organizations over the period of 1982–1990, the period immediately following the establishing of a mainly nonproﬁt system for delivering private health care.1 Although the survey covers a period two decades in the past, it illustrates cross-organizational differences in behavior. At the same time, the Uruguayan health care sector has undergone only very limited reform in the intervening period, leaving the structure of the three organizations, and the regulatory environment, basically unchanged in 2010, although there is signiﬁcant discussion about reform. Taking advantage of this continuity, we conducted interviews with 11 Uruguayan doctors working in the three types of nonproﬁts during the spring of 2010, in order to understand organizational differences more clearly. We show that there are signiﬁcant differences in behavior across organizational forms, even in a context where many behaviors are highly

264

JUAN JOSE BARRIOS AND MIEKE MEURS

regulated. These differences provide partial support for dominant models of nonproﬁt and cooperative behavior. For example, Union (producer) Cooperatives, which are predicted to maximize income per doctor, spend a larger share of total expenses on doctors’ remuneration than other types of organizations, which appear to have different interests. However, some results contradict standard predictions: Mutuals which, as a form of consumer cooperative, are expected to minimize costs to members, do not raise membership fees less than other organizations during a period of price liberalization. Interview evidence suggests some explanations for this and other ﬁndings. We conclude that more attention should be paid to governance structure when making predictions about nonproﬁt behavior.

URUGUAYAN HEALTH SECTOR IN THE 1980s Regulations established in 1981 require all private providers of basic health services to be nonproﬁt organizations (Grau Pe´rez & Lazarov, 1999). There is a long tradition, dating back to the mid-1800s, of nonproﬁt provision of basic health care through Mutuals, mutual beneﬁt organizations based on ethnic solidarity. These organizations already met the new requirements. Other, for-proﬁt, health care providers were required to restructure into nonproﬁts (Bertullo, Isola, Castro & Silveira, 2003, described in Section 3) below. In rural areas, where few nonproﬁts were organized through other channels, the rural doctors’ union (Federacion Medica del Interior, or FEMI) began organizing a minimum of one nonproﬁt provider per departmento (department, or county). By 1984, 36.5% of Uruguayans received primary health care from a nonproﬁt organization (Labadie, 1997), with the share rising to 47% by 1993 (Grau Pe´rez & Lazarov, 1999). The remainder of the population (by 1993 consisting mainly of low-income households without a formal-sector job and families of personnel of the armed forces and police)2 received basic health services from state organizations. Private, for-proﬁt organizations continued to supply specialized (nonprimary care) services. Private, nonproﬁt suppliers of primary care face signiﬁcant regulation. The state sets a minimum number of afﬁliates (members serviced by the health care organization), mandates the package of coverage provided, and sets rates for afﬁliation (membership) and co-payment prices. Regulating the scope of services, the state mandates that organizations supply basic, complete and egalitarian medical assistance in the areas of surgery, pediatrics, gynecology, general medicine, and preventive medicine,

Governance and Behavior in Nonproﬁts

265

as well as other services such as emergency treatment. A health care organization may supply services directly or outsource them to other nonproﬁt health organizations or to private clinics. In this case, the health care organization to which the afﬁliate belongs pays the providers. The system is organized as pre-paid insurance. The state collects a percentage of employees’ wages (from both employers and employees) and pays a membership fee (the ‘‘quota’’) to the health care provider chosen by the individual. If an employee’s contributions do not cover the quota, the state pays the difference, subsidizing health care to the less well-paid. Those afﬁliates who are not formal sector employees but wish to afﬁliate to private nonproﬁt providers must pay the full amount of the state-set quota directly to the medical institution. Health care providers have the right to reject afﬁliates with poor health records, who will then be serviced by public providers. Quotas, the main source of revenue for health care providers, were liberalized to a cap in 1983, and then liberalized completely in July 1984. A rapid increase in the price of membership led to a re-imposition of state-set membership fees in October 1985, however, which was still in place in 1990. Co-payments, the second largest component of health organization revenue, include additional payments for medicines, ofﬁce visits, and other services. Like membership fees, these were brieﬂy liberalized in 1984 and re-regulated in 1985. The remaining income of nonproﬁt health organizations comes from providing services to afﬁliates of other organizations (outsourced services). Since coverage and prices are regulated, competition among nonproﬁt organizations for afﬁliates (and thus revenue) is restricted to things like quality of attention to customers. Competition is further limited by a requirement that individuals afﬁliate to a health care organization located in their own department. Competition is stronger in Montevideo, where in all years of the study there were more than 20 health care nonproﬁts to choose from. In the majority of departments outside Montevideo (10 of 18), only one nonproﬁt provider existed during the entire period under study. In the remaining eight departments, two had three or more providers in every year from 1983 to 1990, whereas the other six had two providers in one to three of the years. Salaries in nonproﬁt health organizations are determined through tripartite negotiations between unions, representatives of the nonproﬁts, and the state. Collective agreements apply to all health organizations in which union members are employed. Doctors need not work more than 26 hours a month for any organization, allowing doctors to work in more than one

266

JUAN JOSE BARRIOS AND MIEKE MEURS

organization. Doctors in fact work in as many as ﬁve or six organizations – often a mix of public organizations, private for-proﬁts, and nonproﬁts. The fact that doctors in different types of organization negotiate their salaries through the same unions apparently leaves little possibility for pay to vary across organizations.3 However, interview evidence suggests that individual organizations are permitted to vary salaries by about 20% from the negotiated rate (Interview 8, 2010). In addition to their salaries, doctors are paid a type of piece rate – an additional sum for each ‘‘medical act’’ that doctors perform. Additional variation in doctor pay may thus emerge across organizations, as these choose the number and type of medical acts performed and deﬁne medical acts, the deﬁnition of which appears to vary across organizations (Interviews 3, 7, 8, 2010). In 1991, payment for medical acts made up about 35% of total remuneration for doctors in Montevideo, and about 30% for doctors in the interior (PAHO, 2007). Finally, doctors may increase their incomes by shifting work – by keeping medical acts ‘‘in-house’’ instead of outsourcing them, or by outsourcing speciﬁc procedures from a nonproﬁt where they work but do not have access to the equipment necessary for the procedure, to another health organization where they also work, where they do have access to the machinery, or to a private for-proﬁt clinic, which they might own (Interviews 2, 7, 9, 11, 2010). Other interview data suggest, however, that in most cases, the decision to outsource services is made by the technical director and is based on relative costs of in-house or outsourced services (Interview 9, 2010). Overall, throughout this period the behavior of Uruguayan health care organizations was highly constrained. They had only brief opportunities to vary the prices they charged for their services, little control over the package of services offered to afﬁliates, and limited ability to inﬂuence remunerations to doctors. Of course, organizations retained more control of some variables, including costs linked to organizational efﬁciencies and bargaining with suppliers, quality of the services to afﬁliates (wait times and doctor quality), and the working conditions of staff. Perhaps surprisingly, investments were not regulated, although they did require state approval.

TYPES OF ORGANIZATION In this section we draw on previous analyses of Uruguayan health organizations as well as interviews with stakeholders to describe the three types of Uruguayan nonproﬁt health care organizations. Eleven interviews (see appendix for details) were carried out with doctors, technical directors,

267

Governance and Behavior in Nonproﬁts

and board members during spring 2010. Since we cannot be sure that the information from the interviews accurately describes condition during the period 1982–1990, we simply highlight ways that interviews suggest differences from the established understanding of organizational structure. In addition to the secondary sources and interview data, we also draw on survey data from the Uruguayan Ministry of Health. The data includes ﬁnancial results, utilization levels, and related information for each of the approximately 50 private, nonproﬁt organizations supplying primary health services over the period 1983–1986 and 1988–19904 (pooled, cross-sectional data). Since many of the organizations did not exist over the entire period, the exact number of organizations varies from year to year, as reﬂected in Table 1. The data was collected retrospectively, through surveys of the nonproﬁt organizations, in 1992. The previous literature (Labadie, 1997; Solari, 1992) describes the three types of nonproﬁt health organizations. Mutuals, described earlier, resemble consumer cooperatives since ‘‘consumers’’ (afﬁliates) elect other afﬁliates to the Board and exercise control over the organization. Doctors working in Mutuals are required to be afﬁliates of the Mutual (Interview 9, 2010). Doctors may be elected to the Board either in their capacity as afﬁliates, or as representatives of the staff. Other employees may also have seats on the Board (as staff), but all staff are minority members. Interview evidence (Interviews 7, 11, 2010) suggests that governance of Mutuals may not always reﬂect broad afﬁliate interests. Board membership Table 1.

Uruguayan Health Care Non proﬁts, by Type and Year, 1983– 1986 and 1988–1990.

Organization Type Year

1983 1984 1985 1986 1988 1989 1990

Mutual

Doctors Cooperatives

CASMU

Union Cooperatives

Total

Number

%

Number

%

Number

%

Number

%

Number

%

23 21 13 13 13 12 12

37 36 28 28 28 24 24

16 15 11 10 10 14 13

26 25 24 22 22 28 27

1 1 1 1 1 1 1

2 2 2 2 2 2 2

22 22 21 22 22 23 23

35 37 46 48 48 46 47

62 59 46 46 46 50 49

100 100 100 100 100 100 100

Source: Uruguayan Ministry of Health (1992).

268

JUAN JOSE BARRIOS AND MIEKE MEURS

is often limited to a small group of insiders, sometimes organized around the families of founding members, who rotate through the Board periodically. Insiders may collude to present a limited number of lists (sometimes only one) at the time of elections, excluding other afﬁliates from power. Boards were also reportedly sometimes ‘‘captured’’ by businessmen-afﬁliates, interested in selling goods and services to the organization. Other health organizations are organized as medical cooperatives, resembling producer cooperatives. These cooperatives can in turn be subdivided into two different types. Guild, or Union, Cooperatives are producer cooperatives organized under the auspices of doctors’ unions. These cooperatives operate independently of the unions in most respects. The board of directors is elected from among doctors who are members of the producer cooperative, by the other member-doctors. In this case, all doctors working in the cooperative are both members and workers. There are two doctors’ unions, one serving doctors working in Montevideo (Sindicato Medico del Uruguay) and the other representing doctors working in other departments (Federacion Medica del Interior, or FEMI). The Montevideo union organizes only one Union Cooperative, CASMU (Centro de Asistencia del Sindicato Medico del Uruguay). CASMU is by far the largest health care organization in Uruguay. There is at least one Union Cooperative organized by FEMI in every department outside Montevideo, organized by the union after the shift to the all nonproﬁt national system in 1981. In some cases, there is more than one in a department. Serving more rural areas, these tend to be much smaller than CASMU. The second type of medical producer cooperative, Doctors’ Cooperatives, are created, and often funded, by doctor-members. Members of the Board are elected from among these doctor-members. In Doctors’ Cooperatives, the member-doctors also hire other, non-member doctors (Labadie, 1997; Solari, 1992). These non-member doctors cannot be elected to the Board.5 Some are regular employees, for whom the organization pays social security contributions. But Doctors’ Cooperatives also reported hiring contract workers. These contract workers may later be offered a chance to become a formal employee (or even a member of the producer cooperative). Although Mutuals are organized by health care consumers, around principles of ethnic or religious solidarity, and then hire doctors, Doctors’ and Union Cooperatives are organized by doctors themselves. The choice to organize as a Union Cooperative, as opposed to a Doctors’ Cooperative, appears to depend in part on when the organization was formed and, relatedly, on who initiated the organization.

Governance and Behavior in Nonproﬁts

269

FEMI took the lead in organizing at least one nonproﬁt health care cooperative in each department after the passage of the 1981 law. This may have been an important service to union members, providing employment to Uruguayan doctors who faced a ﬂooded labor market (PAHO, 2007) but had limited capital for start up costs and other organizational problems. FEMI also appears to offer some other advantages to its Union Cooperatives, such as the ability to ‘‘capture’’ members from other unions, subcontract work from public hospitals (where union members also work), and so on (Interview 4, 2010). Private, for-proﬁt Doctors’ Cooperatives, or group practices, may have simply turned themselves into nonproﬁt a Doctors’ Cooperative at the time of the 1981 law, using capital already in hand. But in rural areas, pre-existing group practices could also work with their union (FEMI) to become a Union Cooperative (Interview 2, 2010). One factor in the choice of organizational form may have been the practice in Doctors’ Cooperatives of hiring non-member doctors as employees. Interviews with members of six Union Cooperatives suggested an important difference from this prevailing description, however. Interviewees explained that (at the time of the interviews in 2010) Union Cooperatives, like Doctors’ Cooperatives, often hire non-member doctors (Interviews 2,3,4, 2010). This ‘‘degeneration’’ reportedly occurs when the founding group of doctor-members can no longer manage the workload, but chooses not to recruit additional doctors as regular cooperative members (Interview 7). It is thus possible that the degeneration occurred later in the Union Cooperatives than in the Doctors’ Cooperatives, and governance differences may still have existed at the time our data was collected. Many Union Cooperatives were new in 1983, and possibly had not yet reached capacity, whereas Doctors’ Cooperatives were more likely to be pre-existing practices transformed after the 1981 law. Uruguayan nonproﬁt health organizations clearly differ in the make-up of their stakeholders and their representation on the board of directors. Boards at Mutuals are dominated by afﬁliates, whereas producer cooperatives’ Boards are dominated by doctors. It is less clear whether the two forms of producer cooperatives are distinct. Other differences among the types of organization that may affect behavior include differences in size, capital stock, services offered, and the age of members. The period 1983–1990 was one of rapid development in the Uruguayan health sector, as for-proﬁt organizations restructured into nonproﬁts and new nonproﬁt ﬁrms were created to meet growing demand. After a 1983 regulation established a minimum number of afﬁliates for health care nonproﬁts, small organizations closed or merged, and the

270

JUAN JOSE BARRIOS AND MIEKE MEURS

number of Mutuals fell. The share of Union Cooperatives in the nonproﬁt population increased as a result, but the number of Mutuals stabilized at about 12, or 24% of the population of nonproﬁts, at the end of the period. As seen in Table 1, the organizations were divided fairly evenly between the three forms. (Because CASMU is uniquely large, we treat it as a distinct type of non-proﬁt, although it is Union Cooperative.) As seen in Table 2, membership in these organizations grew signiﬁcantly over the period 1983–1990. There are signiﬁcant differences in average number of afﬁliates by organizational type however, and these remain fairly stable. As noted earlier, CASMU, the Montevideo Union Cooperative, is signiﬁcantly larger than any other organizational type. Measured by number of afﬁliates, Mutuals are the next largest on average, whereas Doctors’ Cooperatives and Union Cooperatives outside Montevideo are fairly similar in terms of average membership. Over the period, we see a slight shift in membership from Doctors’ Cooperatives to Union Cooperatives. Between 1983 and 1990, capital stock of the nonproﬁts grew much faster than membership (Table 3). There were important differences in the amount of capital acquired, however, with Mutuals and the relatively newer Union Cooperatives gaining capital faster than Doctors’ Cooperatives or the already-capital intensive CASMU. These differences in membership and capital stock may reﬂect different scale economies, particularly affecting the ability of organizations to offer services in-house and to offer (remuneration-enhancing) medical acts to doctors.

Table 2. Uruguayan Health Care Non proﬁts, Average Membership by Organization Type and Year, 1983–1986 and 1988–1990. Organization Type

Mutual

Doctors Cooperatives

CASMU

Union Cooperatives

Year

Mean

Mean

Mean

Mean

1983 1984 1985 1986 1988 1989 1990

22,689 20,580 26,757 31,478 36,287 38,887 41,071

13,583 14,050 18,108 19,501 22,611 16,793 18,305

254,857 250,081 246,543 264,820 272,073 273,512 273,762

12,181 13,031 14,573 16,646 19,751 20,653 21,236

81

35

7

74

% Change

Source: Uruguayan Ministry of Health (1992).

271

Governance and Behavior in Nonproﬁts

Table 3. Uruguayan Health Care Non proﬁts, Average Capital Stock in Constant Pesos, by Organization Type and Year, 1983–1986 and 1988–1990. Organization Type

Mutual

Doctors Cooperatives

CASMU

Union Cooperatives

Year

Mean

Mean

Mean

Mean

1983 1984 1985 1986 1988 1989 1990

49,678 76,497 235,164 420,013 1,270,815 2,458,861 6,266,283

18,811 33,772 99,364 191,707 506,135 708,279 1,686,119

764,391 1,280,509 2,368,795 3,880,892 7,137,517 15,500,000 39,600,000

8,715 16,700 34,861 65,659 210,389 455,308 1,149,680

125

84

51

131

% Change

Source: Uruguayan Ministry of Health (1992).

BEHAVIOR OF NONPROFIT ORGANIZATIONS In this section, we develop hypotheses about behavioral differences across the different types of nonproﬁt organization, drawing on the literature on stakeholder interests and nonproﬁt behavior, and the literature on behavior of producer and consumer cooperatives. Preference differences among diverse stakeholders may create conﬂicts over nonproﬁt objectives. Because nonproﬁts are not subject to take-over threats, and to the extent that they do not face competition from for-proﬁts, the diverse stakeholders make internal organization central to nonproﬁt behavior (Bachiegga & Borzaga, 2002) and generalizations about nonproﬁt behavior difﬁcult. We use information about organizational structure and standard theory of cooperative behavior to develop hypotheses about behavioral differences. Uruguayan health care organizations do not have outside donors, leaving managers, workers, and consumers as relevant stakeholders. Managers may have preferences over the growth and ﬁnancial health of the organization, as a larger, ﬁnancially more successful nonproﬁt may give a manager more power with respect to other nonproﬁts and public ofﬁcials. Managers’ preferences may also include product quality (Glaeser, 2002). Workers in health care non proﬁts include doctors, as well as other medical and nonmedical staff. Their preferences include pay and working conditions. Worker preferences may also include product quality, as nonproﬁt workers

272

JUAN JOSE BARRIOS AND MIEKE MEURS

may choose nonproﬁt employment as a result of their commitment to its goals (DeVaro & Brookshire, 2007; Glaeser, 2002). Since membership fees, co-payments, and the package of beneﬁts are regulated, relevant consumer interests center on quality. The diverse stakeholders may inﬂuence organizational behavior in several ways. The distribution of formal control rights, including membership on the board of directors, is important (Weisbrod, 1975). Control of information and technical knowledge is another channel of inﬂuence. In medical worker cooperatives (hospitals), doctors’ control of technical knowledge may lead hospitals to follow their preferences even when doctors do not constitute a majority of board members (Pauly & Redisch, 1973). Similarly, proximity of certain stakeholders to managers or Board members can increase these stakeholders’ control over important decisions (Glaeser, 2002). Formal control rights in the Uruguayan health care nonproﬁts vary according to their organizational form. In Mutuals, consumers make up the majority of board members, although doctors may also be represented as both consumers and workers, and other workers may also have minority representation on the Board. In Union and Doctors’ Cooperatives, the board of directors is elected from among doctors who are members of the cooperative, by the other member-doctors. An important difference between the two forms of producer cooperatives may be the nonrepresentation of non-member doctors (workers) on the Boards of Doctors’ Cooperatives during the period 1983–1990. Where a Union Cooperative is the only nonproﬁt in a department, doctors will also be consumers in that cooperative, in effect giving consumers some input into decisions. All health care nonproﬁts must have a doctor as their technical director, providing similar informal control rights to doctors across organizational forms – via both legal recognition of doctors’ special knowledge and proximity of the technical director to management and the Board. Other means of inﬂuence via proximity may vary across organizational form. In Mutuals, insider groups of consumer-members may interact closely with the Board and management, while in Doctors’ and Union Cooperatives, senior and founding-member doctors are more likely to do so. Standard models of cooperative behavior assist us in developing predictions about how different types of stakeholder inﬂuence translate into behavior. Although standard models assume a more homogeneous set of stakeholders (workers or consumers) than is actually found in Uruguayan nonproﬁts, they capture the basic differences between the most inﬂuential stakeholders in consumer and producer cooperatives.

Governance and Behavior in Nonproﬁts

273

Many standard predictions of these models will not hold, however, given the highly regulated context of Uruguayan health care cooperatives. For example, a standard model of consumer cooperative behavior (Clark, 1952) posits that cooperatives governed by consumers will minimize per-unit costs. Since the fees and co-payments consumers pay are set by the state, costs to Mutual members are not linked to actual costs of providing the service. Consumer-members may have little incentive to minimize costs. We posit that they may instead maximize quality, by raising doctor remuneration, in order to hire high-quality doctors, by hiring more doctors, to reduce wait time, or by purchasing machines for (convenient) in-house provision of services. Doctors on the Board and the technical director might attempt to inﬂuence consumers’ deﬁnition of quality, to bring this deﬁnition more in line with their own preferences. One important exception to this behavioral prediction, however, would be the period of quota liberalization in 1984–1985. During this period, consumer-members of the Mutual Boards faced a direct link between cooperative behavior (quota-setting) and their costs of health care, and predicted consumer interests in low costs should have affected the behavior of Mutuals. Mutuals would be expected to raise quotas less than other types of organizations. If the capture of Mutual Boards by insiders, described in the interviews, is important, however, the expected cost-limitation would not occur. Standard models of producer cooperatives suggest they maximize net income per member, and as a result choose inefﬁciently high levels of output and investment to achieve this (Domar, 1966; Pauly & Redisch, 1973). In the Uruguayan context, where net income cannot be distributed, the expected behavior might take the form of increases in doctor remunerations and, as in the standard model, greater investment in equipment, which can be used to increase doctors’ remunerations (for medical acts). Producer cooperatives would also be expected to limit outsourcing of services, in order to increase payments for medical acts (unless, as interviews suggest, doctors outsource services to their other places of employment, where they expect higher remuneration). Doctor’s Cooperatives appear to be a ‘‘degenerate’’ form of producer cooperative (Abell, 1983), however, with members hiring non-member workers in a traditional employment relationship. Since hired workers cannot formally inﬂuence decisions, these ‘‘hiring’’ cooperatives may continue to maximize remuneration per member, but this will no longer be equivalent to maximizing remuneration per doctor-producer. The extent to which Doctors’ Cooperatives are expected to behave differently from Union

274

JUAN JOSE BARRIOS AND MIEKE MEURS

Cooperatives depends on their relative dates of degeneration, something that we cannot measure using our data. However, based on descriptions of the development of the sector, we expect a difference in behavior during the period under study. Given the highly regulated context, we focus our analysis on four types of behavior that may vary across organizational type. The ﬁrst three reﬂect differences in overall behavior: remunerations to doctors, investment in medical equipment, and outsourcing. A fourth behavior is the response to the liberalization of quotas 1984–1985. Hypothesized differences with respect to remunerations can be summarized as follows: Union Cooperatives (traditional producer cooperatives) will spend relatively more on doctors’ remunerations than Doctors’ Cooperatives, as the Union Cooperatives seek to maximize remunerations per doctor.6 Doctors’ Cooperatives are expected to maximize revenue per member, which may mean limiting remunerations to non-member doctors and shifting work onto these employees, as member doctors restrict themselves to the best-remunerated medical acts, including those at other places of employment, possibly lowering remunerations to doctors overall. However, if interview evidence that Union and Doctors’ Cooperatives do not differ in their level of ‘‘degeneration’’ applies to the period 1983–1990, no differences are expected in doctors’ remuneration between Union and Doctors’ Cooperatives. Because of uncertainty about how Mutual members deﬁne quality, Mutual behavior is harder to predict. Relative spending on doctors’ remunerations will depend on the extent to which Mutual members deﬁne quality as doctor quality, and whether remunerations are an effective means to attract such doctors. If Mutual members deﬁne quality mainly as doctor quality and higher remunerations attract doctors, Mutuals may not behave differently from Union Cooperatives. If Mutuals deﬁne quality as things like in-house provision of low piece-rates services, however, doctors’ remunerations in Mutuals will differ from those in Union Cooperatives. With respect to investment, Doctors’ Cooperatives are likewise expected to invest relatively less in equipment than Union Cooperatives, investing in equipment needed for medical acts of members, but not investing in machines needed to ‘‘capture’’ medical acts for non-member doctors. However, if Union Cooperatives and Doctors’ Cooperatives do not differ in their level of ‘‘degeneration,’’ no differences are expected in investment between Union and Doctors’ Cooperatives. The behavior of Mutuals will depend, again, on how members deﬁne quality. Mutuals might spend relatively less on investment if it is not needed to attract high-quality doctors or if

Governance and Behavior in Nonproﬁts

275

Table 4. Summary of Hypotheses. H1: Union Cooperatives will spend relatively more on doctors’ pay than other forms of organization H1u : All nonproﬁts will behave similarly H2: Union Cooperatives will spend relatively more on outsourcing than other forms of organization H2u: All nonproﬁts will behave similarly H3: Union Cooperatives will spend relatively more on investments than other forms of organization H3u: All nonproﬁts will behave similarly H4: Mutuals will raise membership fees least during liberalization; Union Cooperative most H4u: Nonproﬁts will respond similarly to liberalization

members do not prefer in-house services (external services may be ‘‘higher quality’’). If Mutuals are captured by insiders, as interviews suggest, this should result in relatively less investment.7 Another set of hypotheses relates to the outsourcing of services. In the Uruguayan context, where doctors are paid by medical act, Union Cooperatives are expected to outsource less than Doctors’ Cooperatives as they seek to maximize remunerations per doctor. Once more, Mutuals’ behavior depends on how quality is deﬁned. They may outsource more, if outsourced services are seen as higher quality. Or they may outsource less, if in-house services are seen as higher quality (perception in-house doctors may use their position to defend). Finally, we examine whether organization types respond differently to the liberalization of quotas in 1984. We expect that Union Cooperatives, maximizing remunerations, will raise quotas most, whereas Mutuals, where consumers are majority Board members, will raise quotas least. Doctors’ Cooperatives, unconstrained by members, are expected to behave more like Union Cooperatives, more if they are ‘‘degenerate.’’ Hypotheses are summarized in Table 4.

ANALYSIS OF THE URUGUAYAN CASE In examining differences in doctors’ remunerations, investment, and outsourcing of services, we normalize these expenditures as a share of total expenses. Because this risks introducing other, distorting, factors into our

276

JUAN JOSE BARRIOS AND MIEKE MEURS

outcome variable – those affecting total expenses – we also did all of the analyses normalizing by number of afﬁliates, as a robustness check. Other than slightly reducing the signiﬁcance of the regressions, the use of the spending-per-afﬁliate measures does not change the results.8 Table 5 shows variation in the outcome variables across organizational type. As expected, Union Cooperatives and CASMU spend a larger share of total expenses on payments to doctors than Mutuals or Doctors’ Cooperatives. As expected, Union Cooperatives also spend the least on outsourced services. However, for investment, we do not see the expected differences. Union Cooperatives (other than CASMU) have lower average levels of investment than Doctors’ Cooperatives or Mutuals. Surprisingly, we see quite a bit of variation in quotas per member, although for most of the period quotas were set by the state. Fig. 1 shows that this is not simply a result of differing responses to the liberalization in 1984–1985. Differences exist in every year, perhaps due to noise in membership data or a discrepancy between the dates that quotas are paid and membership is counted. Importantly, Fig. 1 shows differences in the change in quotas per member during the 1984–1985 liberalization, especially in 1985, before the re-regulation, which was announced in advance. In the regression analysis, we control for variables other than organizational form which might affect behavior. All variables are annual reported values for each organization. We control for size of organization (number of

Table 5. Remunerations, Outsourcing and Investment as % 1986 Total Expenses, and Quota per Member (in ‘000 Pesos) by Organization Type and Year, 1983–1986 and 1988–1990. Type

Mutuals Doctors Cooperatives CASMU Union Cooperatives Total

Doctors’ Remunerations

Investment

Outsourced Services

Quota per Member

Mean Standard Deviation

Mean Standard Deviation

Mean Standard Deviation

Mean Standard Deviation

0.30 0.28

0.08 0.08

3.65 4.67

0..20 0.23

0.20 0.23

0.11 0.11

2.82 2.59

0.84 0.74

0.37 0.42

0.05 0.07

5.24 2.75

0.16 0.15

0.16 0.15

0.02 0.08

3.17 2.63

0.43 0.41

0.35

0.10

3.41

0.46

0.46

0.10

2.69

0.66

Source: Uruguayan Ministry of Health (1992).

Governance and Behavior in Nonproﬁts

Fig. 1.

277

Quotas per Member, by Organization Type and Year, 1983–1986 and 1988–1990.

afﬁliates). Size may measure economies of scale in in-house service provision and affect outsourcing and also investment. Similarly, we control for capital stock (ﬁxed assets, in constant pesos). Different types of organization face different competitive environments. CASMU is located in Montevideo, so faces signiﬁcant competition. Doctors’ Cooperatives and Mutuals are never monopolies, but Union Cooperatives often are. These differences in competitive environment may affect organization behavior, and we control for this in two ways. To control for possible competition for afﬁliates, we include the number of other health care organizations operating in the same department. This varies from 0 in every year in many departments to over 30 in the early years in Montevideo. As a control for competition for doctors (the labor market for which faces a signiﬁcant surplus in Montevideo but may face a shortage in less desirable areas), we include a dummy variable taking a value of 1 for Montevideo and the neighboring departments of Canelones and San Jose, and a 0 for all other departments.9 The need for doctor visits and other services may increase with age, and some of the older organizations may have older afﬁliate populations. We control for the share of afﬁliates older than 65 years. Finally, the sector was evolving in signiﬁcant ways over the period 1983–1990, as an increasing share of the rural population was incorporated into the system, new organizations entered and matured, and smaller

278

JUAN JOSE BARRIOS AND MIEKE MEURS

organizations were eliminated. To control for this, as well as cyclical economic impacts, we include a complete set of year dummies, with the ﬁrst year, 1983, as the omitted category. When comparing the way the 1984–1985 liberalization affects quota-setting across organizational types, we replace the full set of year dummies with an interactive variable. We interact organization type with a dummy for the year 1985 (the year of most change in quotas, seen in Fig. 1). Our variable of interest, organizational form, is measured as a set of dummy variables, with CASMU, the very large Montevideo Union Cooperative, as the omitted category. Means and standard deviations of these variables are presented in Table 6. The econometric model takes the following form: yit ¼ axit þ bzit þ where yit represents, alternatively, the four outcome variables used in the four regressions: Doctors’ incomes, spending on outsourced services, and investment, as a share of current expenses, by organization, at time t, and change in quotas per member over the previous year, in thousands of constant pesos, by organization i, at time t, all for t ¼ 1983–1986 and 1988–1990. Because we estimate investment and change in quota per member using the year-on-year change in capital stock and quotas, these regressions do not include 1983, and 1984 becomes the omitted category for the year dummy. Further, we use the natural log of the investment share variable in the regression, in response to the extreme distribution of this variable (which ranges from 0.51 to 75.52).

Table 6.

Statistical Characteristics of Explanatory Variables.

Variable Share older members Capital Stock Year 1990 Year 1989 Year 1988 Year 1986 Year 1985 Year 1984 Number of competitors Montevideo region Number of afﬁliates

Mean

Standard Deviation

0.112 824891.100 0.137 0.137 0.128 0.128 0.128 0.165 11.904 0.571 25554.530

0.091 3131471.000 0.344 0.344 0.335 0.335 0.335 0.371 12.198 0.495 39531.010

Source: Uruguayan Ministry of Health (1992).

Governance and Behavior in Nonproﬁts

279

xit represents the set of dummy variables for organization type for each organization at time t, although no organization switches type between years; CASMU, the very large cooperative of the Montevideo doctors’ union, is omitted, and zit represents the other explanatory variables for each organization at time t. Capital stock and number of afﬁliates are clearly affected by the behavioral outcomes examined here – investment in one period affects the capital stock in the next, and number of afﬁliates will be affected by organization spending on doctors and investment. Here, we include capital stock and afﬁliates as independent variables to control for the impact of current levels on current year behavior. We estimate the regressions as Robust OLS regressions in STATA.10 The regression results are presented in Tables 7 and 8. We begin by discussing Table 7. The ﬁndings conﬁrm that, controlling for differences in size, competitive environment, capital stock, patient demographic, and year, there are signiﬁcant behavioral differences across organization type. One important ﬁnding is that CASMU, the omitted category of organization, behaves differently than any other type of organization. Compared to all other types of health care nonproﬁt, CASMU spends a lower share of its expenses on doctors’ pay and investment, and a greater share on outsourcing. Although we have attempted to control for the important differences in size and the competitive characteristics of location, perhaps these controls are imperfect. Because CASMU is a single cooperative, located in Montevideo, the CASMU organizational variable may be picking up speciﬁc characteristics of this location, in addition to organizational differences. CASMU has many possibilities for outsourcing, including to other clinics where CASMU doctors work, a practice mentioned by interviewees. This consideration may also drive decisions about investment. Desirability of employment in this large, well-located and well-endowed organization may limit the need to raise doctors’ salaries to attract and retain doctormembers. All three other organizational types spent a signiﬁcantly larger share of expenses on doctors’ remunerations than CASMU. As expected, spending in Union Cooperatives, which are predicted to maximize remunerations per doctor, was higher than in other forms. (The coefﬁcient on Union Cooperatives is signiﬁcantly larger than that on Doctors’ Cooperatives and Mutuals at a level of po0.001.) However, spending on doctors’ remunerations did not differ signiﬁcantly between Mutuals and Doctors’ Cooperatives. As suggested earlier, Mutuals may not seek to reduce this cost, preferring to spend more to attract high-quality doctors. Spending on doctors’ remunerations was negatively

280

JUAN JOSE BARRIOS AND MIEKE MEURS

Table 7.

Regression Results. Behavioral Differences across Health Organizations.

Dependent Variables Explanatory Variables Mutuals Doctors Cooperatives Union Cooperatives Number of afﬁliates Capital stock Number of competitors Montevideo region Share older members Year 1983 Year 1984 Year 1985 Year 1986 Year 1988 Year 1989 Year 1990 Constant Observations R2 Robust standard errors in parentheses

Remunerations

Outsourcing

ln Investment

0.116 (0.035) 0.108 (0.357) 0.237 (0.0540 0.000 (0.000) 0.000 (0.000) 0.001 (0.001) 0.016 (0.014) 0.097 (0.066) Omitted 0.012 (0.012) 0.009 (0.014) 0.008 (0.015) 0.003 (0.015) 0.014 (0.016) 0.015 (0.016) 0.157 (0.045) 338 0.507

0.319 (0.040) 0.309 (0.047) 0.385 (0.050) 0.000 (0.000) 0.000 (0.000) 0.002 (0.001) 0.035 (0.014) 0.162 (0.094) Omitted 0.009 (0.018) 0.010 (0.019) 0.008 (0.018) 0.016 (0.019) 0.018 (0.019) 0.020 (0.020) 0.578 (0.056) 338 0.236

7.072 (1.164) 7.203 (1.250) 8.197 (1.283) 0.000 (0.000) 0.000 (0.000) 0.003 (0.028) 1.106 (0.276) 3.468 (1.192) – Omitted 1.060 (0.300) 1.573 (0.290) 2.822 (0.337) 3.383 (0.330) 4.201 (0.310) 1.079 (1.411) 263 0.717

po0.001, po0.05, and po0.10.

correlated with the number of competitors, as expected. Location in the Montevideo region was not signiﬁcantly related to the share of doctors’ remunerations in total expenses, except to the extent that the CASMU dummy picks up this effect.

281

Governance and Behavior in Nonproﬁts

Table 8.

Regression Results. Differential Response to Liberalization across Health Organizations.

Dependent Variable Explanatory Variables Mutuals 1985 Doctors Cooperatives 1985 Union Cooperatives 1985 Number afﬁliates Capital stock Number competitors Montevideo region Share older members Year 1984 Year 1986 Year 1988 Year 1989 Year 1990 Constant Observations R2 Robust standard errors in parentheses

Change in Quota per Member 0.587 (0.138) 0.362 (0.139) 0.693 (0.111) 0.000 (0.000) 0.000 (0.000) 0.008 (0.004) 0.083 (0.084) 0.111 (0.312) Omitted 0.075 (0.084) 0.077 (0.084) 0.029 (0.085) 0.018 (0.090) 0.053 (0.075) 280 0.209

pW0.001, pW0.05, and pW0.10.

Looking at outsourcing of services, Union Cooperatives (other than CASMU) again behave as expected. Union Cooperatives spend a signiﬁcantly lower share of expenses on outsourcing of services than Mutuals and Doctors’ Cooperatives (the coefﬁcient on Union Cooperatives differs with a probability pW0.01), perhaps in order to increase payments

282

JUAN JOSE BARRIOS AND MIEKE MEURS

for in-house medical acts to member-doctors. Again, the small difference in the coefﬁcients on the dummies for Mutuals and Doctors’ Cooperatives is insigniﬁcant. After controlling for the behavior of CASMU, location in the Montevideo region is negatively related to outsourcing. Surprisingly, capital stock does not affect outsourcing, controlling for other factors. Organizations with more members and with a greater share of members over 65 years outsource relatively less.11 Investment behavior also follows the expected pattern of signiﬁcantly higher levels of spending by the Union Cooperatives, compared to Mutuals or Doctors’ Cooperatives (po0.01). Spending more on investment may permit more medical acts by doctor-members. As with other behavior, the small difference between the coefﬁcients on the dummy variables for Mutuals and Doctors’ Cooperatives is not signiﬁcant. Organizations with more afﬁliates spend more, although the effect is very small. Being in the Montevideo region and having a greater share of older members are also associated with more spending on investment. Year is signiﬁcantly related to relative spending on investment, with a steady rise in the size of the coefﬁcients over the entire period. Finally, we consider whether the different types of organization behaved differently during the brief liberalization of quotas (Table 8). In this regression, we replace the year dummy for 1985 with an interactive variable, interacting 1985 with organization type. As might be expected, CASMU again behaves differently from all other organizational types, raising quotas the least. Among other forms of organizations, Union Cooperatives raise quotas the most, as predicted. The coefﬁcient on Union Cooperatives is signiﬁcantly different from that on Doctors’ Cooperatives and Mutuals (po0.10 and o0.05, respectively). Contrary to expectations, however, Mutuals do not raise quotas the least. The coefﬁcient on the dummy for Mutuals is not signiﬁcantly different from that on Doctors’ Cooperatives. This suggests that the behavior of Mutuals may not be driven by consumers’ interests. Perhaps, as suggested by interviews, the Boards governing Mutuals are often captured by insiders interested in selling (relatively high cost) services or materials to the cooperative, thus preferring high quotas to cover such costs.

CONCLUSIONS Using data on Uruguayan health organizations for the period 1983–1986 and 1988–1990, supplemented with interviews, we have examined the relationship between governance structures and behavior in nonproﬁt health

Governance and Behavior in Nonproﬁts

283

care cooperatives. We ﬁnd some evidence that different types of nonproﬁts behave differently, despite a tightly regulated environment with prices and wages set externally. As suggested by the literature on nonproﬁt and cooperative behavior Union (producer) Cooperatives spend a larger share of total expenses on doctors’ remuneration than Doctors’ Cooperatives (a possibly more degenerated form of producer cooperative) and Mutuals, where consumers have broader interests, beyond raising doctors’ pay. In some cases, the theory on cooperative behavior does not predict outcomes, however. Mutuals, which are predicted to further consumer interests, did not raise membership fees less than other forms of organization during a brief price liberalization. Interview evidence, which provides additional information on governance structure, suggests that the behavior of Mutuals may be driven more by the interests of small groups of insiders than by those of consumer-members as a whole. Another insight from the interviews, the possible importance of hired doctors in Union Cooperatives, suggested behavioral commonalities with Doctors’ Cooperatives not seen in the data. This may be due to changes between the period of the survey and the time of the interviews (a gradual degeneration of Union Cooperatives since 1990), or perhaps the small interview sample of Union Cooperatives, where hiring non-member doctors was reported to be common, is not representative of Union Cooperatives as a whole. The difference in behavior across organizations with differing governance structures provides support for recent efforts to develop more ﬁne-grained predictions of behavior of nontraditional ﬁrms (Ben-Ner & Gui, 2003). It also suggests that empirical tests of behavioral differences between traditional and nontraditional ﬁrms will be more meaningful if the governance structure of nontraditional ﬁrms is common, and clearly speciﬁed. An important limitation of this study is our inability to control for the exact timing of degeneration of producer cooperatives, and this would be one important element of governance structure to consider in future data collection.

NOTES 1. The database was provided by Gaston Labadie who studied Uruguayans’ choice among various types of health care organisations (Labadie, 1997). Labadie received the data from the Uruguayan Ministry of Health. 2. Additional public services are offered through the hospital of the Banco de Seguros del Estado (State Insurance Company) for certain health issues, such as

284

JUAN JOSE BARRIOS AND MIEKE MEURS

individuals seriously burned, and the Fondo Nacional de Recursos (National Resource Fund), a decentralized publicly funded organization that funds high-cost, complex medical treatments, often through private providers. 3. As will be discussed further later, doctors in Montevideo (regardless of the type of organization they work in) are organized by a different union than doctors in other provinces and, facing a much larger supply of doctors in Montevideo, members of this union typically get somewhat lower salaries than members of the union which organizes doctors outside Montevideo. 4. The data is meant to be a complete series from 1982 to 1900. However, data for 1982 and 1987 are missing. 5. Both types of producer cooperatives have rules which allow non-member doctors, workers and afﬁliates to be represented on the Board, but this possibility has not been exercised in practice. 6. Where Union Cooperatives are the only nonproﬁt health care provider in a department, the doctor-members will also be consumers. This will not be true for all Union Cooperatives. However, it may have an impact on behavior if doctorconsumers deﬁne quality as doctor quality. 7. Examples of capture had to do mainly with contracts for consumables. 8. Results available from authors. 9. A Montevideo dummy might be preferable for capturing speciﬁc characteristics of this labor market (as will be discussed later). However, CASMU, a unique organization type, is located only in Montevideo, creating a collinearity problem with a Montevideo dummy. Dropping CASMU from the sample is an alternative strategy. Dropping CASMU does not affect results regarding the behavior of other organizational types, so we include CASMU in the sample. 10. The Robust OLS regression in STATA uses Cook’s D to select (and then drop) outliers. The use of the Robust Regression produces almost identical results to the standard OLS regression, but results in slightly higher levels of signiﬁcance because of improved standard errors. 11. This somewhat surprising result may appear linked to overall higher costs, which would increase the denominator of the dependent variable. However, as noted earlier, this result also holds when measuring outsourcing as expenditure per member.

ACKNOWLEDGMENTS The authors would like to thank the US Fulbright Commission for support for this project, Ivanova Reyes for research assistance, and two anonymous reviewers for helpful comments. For all remaining limitations of the chapter, the authors take full responsibility.

REFERENCES Abell, P. (1983). The viability of industrial producer co-operation. In: C. Crouch & F. Heller (Eds.), International yearbook of organizational democracy (Vol. 1). Chichester: Wiley.

Governance and Behavior in Nonproﬁts

285

Bachiegga, A., & Borzaga, C. (2002). The economics of the third sector. In: H. K. Anheier & A. Ben-Ner (Eds.), The study of the nonproﬁt enterprise, theories and approaches. New York: Kluwer Academic/Plentum Publishers. Ben-Ner, A., & Gui, B. (2003). The theory of nonproﬁt organizations revisited. In: H. K. Anheier & A. Ben-Ner (Eds.), The study of the nonproﬁt enterprise, theories and approaches. New York: Kluwer Academic/Plentum Publishers. Bertullo, J., Isola, G., Castro, D., & Silveira, M. (2003). El Cooperativismo en Uruguay. Retrieved from http://www.universidadur.edu.uy/bibliotecas/trabajos_rectorado/ doc_tr22.pdf. Clark, E. (1952). Farmer cooperatives and economic welfare. Journal of Farm Economics, 34(1), 35–51. DeVaro, J., & Brookshire, D. (2007). Promotions and incentives in nonproﬁt and for-proﬁt organizations. Industrial and Labor Relations Review, 60(3), 311–339. Domar, E. (1966). The Soviet collective farm as a producer cooperative. American Economic Review, 56, 734–757. Fizel, J., & Nunnikhoven, T. (1992). Technical efﬁciency of for-proﬁt and non-proﬁt nursing homes. Managerial and Decision Economics, 13(5), 429–439. Glaeser, E. (2002). The governance of not-for-proﬁt ﬁrms. Discussion paper no. 1954. Harvard Institute of Economic Research, Harvard University, Cambridge, MA. Grau Pe´rez, & C., Lazarov, L. (1999). Sistemas de salud, y previsio´n social en Uruguay. Banco de Previsio´n Social discussion paper. Montevideo, Uruguay. Interviews, Uruguayan Health Care Organizations, Uruguay, Spring 2010. Labadie, G. (1997). Regulacio´n y Desempen˜o Comparado de Dos Subsistemas Privados de Salud en el Uruguay. Serie de Documentos de Trabajo R-307. InterAmerican Development Bank. Washington, DC. Mehdi, F., & Filippini, M. (2004). An empirical analysis of cost efﬁciency in non-proﬁt and public nursing homes. Annals of Public and Co-Operative Economy, 75(3), 339–365. Pan American Health Organization. (2007). Perﬁl de recursos humanos del sector salud en Uruguay. Retrieved from http://www.new.paho.org. Pauly, M., & Redisch, M. (1973). The not-for-proﬁt hospital as a physicians’ cooperative. American Economic Review, 63(1), 87–99. Peters, L. L. (1993). Non-proﬁt and for-proﬁt electric utilities in the United States; pricing and efﬁciency. Annals of Public and Co-Operative Economy, 64(4), 575–604. Solari, A. (1992). Asistencia Me´dica Colectiva: Formas de Organizacio´n y Marco Normativo. Centro de Estudios de la Realidad Econo´mica y Social (CERES). Working paper No. 10, Montevideo, Uruguay. Uruguayan Ministry of Health (1992). Weisbrod, B. (1975). Towards a theory of the voluntary nonproﬁt sector in a three sector economy. In: E. S. Phelps (Ed.), Altruism, morality and economic theory (pp. 171–195). New York: Russell Sage Foundation.

286

JUAN JOSE BARRIOS AND MIEKE MEURS

APPENDIX. LIST OF INTERVIEWS (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11)

Montevideo Doctors’ Cooperative, Doctor Colonia, Union Cooperative, Doctor Colonia, Union Cooperative, Former Technical Director San Juan, Union Cooperative, Technical Director Montevideo, CASMU, Doctor Montevideo, CASMU, Doctor, CEO Montevideo, Mutual, Doctor Montevideo, Doctors’ Cooperative, Doctor, Technical Director Montevideo, Mutual, Members of Board, CEO Montevideo, Doctors’ Cooperative, Doctor Montevideo, Mutual, Doctor

PART IV FREE TRADE AND THE ECOLOGICAL EFFECTS OF ALTERNATIVE SOCIO-ECONOMIC SYSTEMS

CAPITALISM, ECONOMIC DEMOCRACY, AND ECOLOGICAL DESTRUCTION OF OUR PLANET Jaroslav Vanek ABSTRACT There is a fundamental difference between the impacts of two alternative systems of economic organization: capitalist or fully democratic. The latter, based on democratic decisions based on personal rights, including in the area of enterprise management and organization will, in many contexts, protect the natural environment because the decision makers live in and are permanently exposed to that environment. By contrast, the capitalist ﬁrm and the system based on it and on proﬁt maximization (where the often ‘‘atomized’’ owners have never even seen their ﬁrm) will tend to avoid where possible all environmental-related costs, and thus hurt the natural and human environment. Thus public regulation of capitalist ﬁrms will be far more called for than in the case of economic and full democracy. In the chapter that follows I make an attempt to substantiate these claims. Keywords: Performance and prospects; comparative analysis of economic systems JEL classiﬁcation: P 51

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 289–298 Copyright r 2011 by Emerald Group Publishing Ltd All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012015

289

290

JAROSLAV VANEK

INTRODUCTION Probably the most signiﬁcant issue affecting the consciousness of all people, whether in the rich or the poor parts of the world, is the ecological impact on – or gradual destruction of – our planet. The purpose of this study is to examine this issue from a rather new and unusual point of view, that of socioeconomic systems. To be more precise and to clarify our argument, it is desirable to deﬁne more carefully what we mean by ecological impact. People and in general human societies by their various activities produce what we refer to as ‘‘ecological impact’’ on the world we live in. What concerns us here is such impact bearing on the wellbeing of humankind. It can be positive or negative, and it is the second that will concern us here, as much as the concern of all those who are considering the ecological issue. The typical subject under the ‘‘ecological’’ heading includes such things as global climate change and its effect on people, such as the effect of global warming on the life of people in low-lying parts of the planet or impact of global warming on loss of rain, irrigation, and food production. The typical reaction of world governments and institutions is the attempt to identify precisely the extent of such impacts, and where possible to cope through policy or otherwise with the negative effects or impact of such ecological developments, such as the Kyoto agreements or the meetings at the end of 2009 in Copenhagen. All this is well known, and experts and politicians better qualiﬁed than this writer are or should be dealing with the subject. But there is a related issue very rarely if ever noted today that should also interest us and that may be quite critical for the ecological strategy of the world, and even of vital signiﬁcance in the long run for the welfare of humanity. This topic is the socio-economic system under which the world economies or countries are operating. A Stalinist or soviet-type system certainly would have had its – probably most negative – effect on the world ecology. A fully monopolistic capitalism, if it had not been mitigated by corrective antitrust legislations and action, would have had its own effects: The White House and the whole world economy would have most likely been ‘‘owned by the Rockefellers’’ but the world ecology under a benevolent dictator might have been better because the ‘‘dictator’’ would have preferred cleaner air for himself and his friends. The system of regulated oligopolistic competition under which the world is (more or less) living under today has its own effects on world ecology –

Capitalism, Economic Democracy, and Ecological Destruction of Our Planet

291

and we know that the present-day state of the world ecology is far from desirable. And it is precisely the policy action attempted by the world governments that is aimed at corrective action. But there is another economic system not as well known but certainly well understood in both theory and concrete observation that is likely to yield better results for the world ecology. It is the system of what we refer to as economic democracy – or full democracy where all aspects of social existence, including the sphere of production, are based on democratic principles. By this we understand a system of one person one vote by those involved; instead of one person one vote in the political sphere and one dollar one vote in the rest of human endeavor. It is this comparison, as much as we can, that we want to endeavor in, between the systems of oligopolistic capitalism and a system of economic democracy, both operating under free competitive market conditions that we want to study here from the point of view of their respective ecological impacts on life on our planet. We will have a lot to say, both in theory and in practice, in the sections below. But perhaps one very concrete example can illustrate what we understand by such a comparison. Take the well-known and in some sense tragic situation where most transportable products are produced in China because of the more than one thousand percent wage differential. Crudely speaking, the oligopolistic capitalist solution is to close factories and eliminate jobs in the United States and produce in China for the American markets. This leads to misery for not only our unemployed workers but also ten thousand miles worth of transportation and fuel consumption, not to mention corruption and income maldistribution in the world economy. Contrast this with the ‘‘theoretical’’ fact that the workers of American factories cannot move to China, but if they could control democratically their living and working conditions, they could stay home and seek solutions – which we will speak about presently – which avoid both the human and the ecological negative impacts.

THE THEORETICAL ANALYSIS For the purpose of theoretical analysis we ought to start with a more formal deﬁnition of the two systems we want to consider. The capitalist system or economy is usually deﬁned as a market economy where individual ﬁrms are motivated by the desire to maximize proﬁts.

292

JAROSLAV VANEK

The proﬁt to be maximized is deﬁned as revenue (R) minus labor costs (L) minus all other costs (O). But maximization of R implies minimization of L and O. To the extent that the managers of the capitalist ﬁrms can minimize L or O they will do it. And in both situations such action will or may have a negative ecological impact. We cannot go into the detail of such actions, but we know that this can lead to signiﬁcant global losses of wellbeing. The usual argument given by the apologists of capitalism is that competitive labor and product markets give the managers constant market prices. But we also know that in many cases – cases that can be enumerated by the reader in thousands – such conditions do not apply, as in the case of dumping lethal chemicals into public waterways or into public grounds; or the moving of production abroad (outsourcing) while throwing labor out of jobs. In fact, if we include into our analysis also the uncompensated use of polluting energy, all these impacts of the capitalist system can be subsumed as belonging to the ecological destruction of the planet. Perhaps more concretely and speciﬁcally the negative impact on L by virtue of the action of proﬁt-maximizing capitalism can be referred to as the minus sign syndrome (MSS). But there are other aspects of modern capitalist organization which further strengthen the comparative negative ecological impact. In the present context of ecological impact, a self-employed or family enterprise would appear as much better, because ﬁrst of all the MSS is not present where the self employed producer maximizes his or his family’s wellbeing – which includes scores of objectives. And some of these objectives can control the destructive impacts on the environment, because the worker or his family live in or very near the environment affected. The democratic or self-managed ﬁrm is not ideal, but in many respects it resembles the individual producer. Above all, it also is not subject to the MSS because it does not maximize proﬁt but also maximizes the well-being of its members, subject to some democratically determined principles of overall net income distribution. Also, since the working members of the ﬁrm live within the natural environment of the enterprise, they will tend to protect that environment for their own beneﬁt. Some of the negative ecological impact may of course remain, but the overall positive tendency cannot be denied for the democratic ﬁrm. But there are other and equally signiﬁcant comparative advantages of the democratic forms of ﬁrms and industries, in that they are less or not wasteful of resources of many kinds. This thesis takes a little more explanation.

Capitalism, Economic Democracy, and Ecological Destruction of Our Planet

293

Given these objectives of the two types of economies and given technologies – for the moment assumed to be identical for both – we must begin our inquiry by studying the respective equilibria (or solutions) of ﬁrms and industries. The typical technology as we encounter it in the real world is one subject to increasing returns to scale, followed by constant returns to scale or even diminishing returns for very large scales of production. Given such technologies and given competitive markets, the capitalist ﬁrm will grow to a size beyond the scale of constant returns because at such returns the managers would always try to increase proﬁts by increasing level of outputs. And if, as is most often the case, the range of constant returns is large or even inﬁnite, the ﬁrm would tend to grow indeﬁnitely up to a point where the demand becomes less than inﬁnitely elastic, and the ﬁrm tends to become monopolistic. This is precisely the situation that we described by the reductio ad absurdum of the ‘‘Rockefeller White House.’’ The negation of the absurd solution then is sought through antitrust legislation. But that solution then leads to another and perhaps worse can of worms through creating a situation of oligopolistic competition. Thereby a few ﬁrms compete through high pressure advertising and increasing demand and use of resources beyond what is reasonable, while ‘‘brainwashing’’ the population and turning its consciousness to vulgar materialism and seeking ever newer products – instead of life enjoyment through more moral and ecological endeavors. By contrast, the solution of a democratic ﬁrm and industry is far more ecological in many respects. First of all the democratic ﬁrm maximizing income per member or per worker will ﬁnd its equilibrium at the point of maximal efﬁciency of production factors – not growing beyond that point and leading to monopolistic results (see my General Theory of Labor Managed Economies) and not calling for antitrust or oligopolistic results. Moreover the democratic solution, thanks to relatively smaller and less aggressive ﬁrms, will tend to avoid the ecologically undesirable solutions just noted for the capitalist oligopolistic solution. In addition to using a formal analysis of sales promotion and advertising, ceteris paribus, the capitalist ﬁrm will tend to use more wasteful advertising (pushing its products) than the democratic ﬁrm that may not even use such techniques. One of the signiﬁcant factors of the democratic ﬁrm adjusting to optimal allocation of its resources is that it can control its working hours. This is most likely to lead to reduced outputs as compared to a capitalist situation, thus reducing the pressure on resources including energy consumption. The capitalist to maximize proﬁts must increase or control at a high level the

294

JAROSLAV VANEK

labor inputs because this is required for his proﬁt; whereas the working community can maximize its welfare through taking out some of its satisfaction through leisure, especially since the labor incomes per hour can be higher than the capitalist wage. Using a somewhat similar train of thought, it is possible to expect a difference between the two economic solutions, favoring the democratic ﬁrm with respect to the quality of the product – better quality requiring often less material or even less energy. The self-determining working community will often choose or prefer more precise and enjoyable effort; whereas the capitalist will often introduce brain-wrecking and physically demanding work on the assembly line of inferior products with repetitive action at high speed. But such production at high volume of lesser quality goods will call for more material and energy per unit of output. A reductio ad absurdum may illustrate our argument: It might be most attractive and agreeable to the worker to produce every part of an automobile entirely ‘‘by hand.’’ But such a less stressful solution could hardly be acceptable in modern manufacturing. One of the most signiﬁcant causes of ecological deterioration is transportation and its related consumption of fuel that pollutes the environment. And thus if there are systemic causes of reduction of transportation through reduced distances, there is a deﬁnite result of ecological improvement. There are several reasons to expect such positive effects from an economy based on fully democratic principles, not only political, but also and especially in the organization of production. The ﬁrst positive effect we noted already in the introduction related to the democratic creation of ‘‘offspring’’ production units by self-managing or worker-owned ﬁrms. By contrast to the capitalist ﬁrm that shuts down in the United States and moves to China, the democratic ﬁrm – as illustrated by the cooperatives of Mondragon – preserves the jobs at home in Spain and even increases the economic performance by creating offspring in some other parts of the world with signiﬁcantly lower incomes and needed employment. This policy can have many ecological results. Instead of thousands of miles to and from China, the distances involved by the offspring ﬁrm can be reduced. There can occur a division of labor between more and less complex products between the mother ﬁrm and the offspring. Often also the offspring ﬁrm is located closer to the loci of consumption or demand; thus the transportation is signiﬁcantly reduced. Another reduction of transportation effect is likely to occur through an increased number of competitive ﬁrms, smaller in equilibrium than the capitalist oligopolistic ﬁrm. A theoretical example will illustrate the point: suppose that a market in the United States is supplied by a single

Capitalism, Economic Democracy, and Ecological Destruction of Our Planet

295

oligopolistic ﬁrm from the center of the country, in Chicago. If the market is supplied more locally by a number N of democratic ﬁrms evenly distributed around the country, and the transportation distance of the capitalist centralized alternative is say, a trillion miles (from factory to equally distributed consumers/users) the transportation aggregate distance from the N democratic ﬁrms equally distributed around the country will be equal to a trillion divided by square root of N. For a more precise derivation of this result see Chapter 25 of my Uniﬁed Theory of Social Systems (available at http://hdl.handle.net/1813/642). In fact, this theoretical result has a very signiﬁcant concrete manifestation in the modern development of thousands of local farmers markets, where a single supplier or a supplier region in California is replaced by thousands of local growers selling in farmers markets. The reduction of distances may not be exactly the ‘‘square-root formula’’ but the formula is a good representation of the orders of magnitude of transport reduction and corresponding ecological improvement. In addition, the local farm producers are often very small one family or other democratic ﬁrms, which also save jobs and create a more interesting and creative type of work. We have referred in several contexts to the inefﬁciency of the capitalist system with respect to transportation over long distances and the corresponding overuse of polluting fuels. One of these was the case of the alternative of offspring ﬁrms of cooperative and democratic ﬁrms. But this is only a minor portion of our argument about destructive trade. By destructive trade, we understand a double destruction; one of jobs in the United States, and second the ecological destruction of the environment through the wasteful use of energy in transportation over long distances, such as importing from China trillions of dollars of goods that could have been produced in the United States. But there is a hidden inefﬁciency and destruction – similar to the mental destruction through high pressure advertising/pushing – namely the destruction of scientiﬁc minds that prefer a theory of superiority of free trade, which is false and which is accepted by capitalist economists because it serves the capitalist leaders who beneﬁt from free trade while hurting the workers and the rest of the population. The fallacy of the free trade argument is explained in a short paper included as an Appendix to this chapter. The destructive trade analysis and a corresponding discussion of remedial solutions can be found in my less known study published by the Tinbergen Institute in India. (‘‘Destructive International Trade from Justice for Labour to Global Strategy,’’ International Journal of Development Planning Literature, 13(1), 1998).

296

JAROSLAV VANEK

Related to this discussion and also relying on the analysis of the appendix is perhaps the most powerful argument to be introduced here in concluding our discussion: Relying on the erroneous theory of free trade – or using the lying thereof – in the United States and perhaps other developed countries, we have moved most of our manufacturing capacities to countries like China. Thereby we have stimulated if not originated the superdevelopment of that country, which now is becoming the dominant polluter of the planet. And thus for the 30 pieces of capitalist proﬁt – here and in China – we have sold down the drain the welfare and happiness of millions of working families in the United States. In addition we have strengthened the forces of ecological destruction in the whole world.

APPENDIX: REVISION OF THE THEORY OF COMPARATIVE ADVANTAGE AND EXPLANATION OF THE 2008 WORLD CRISIS I will be very brief because there is a lot to say and my arguments are of considerable relevance for understanding and analyzing the present world economic crisis. For easy reference, I number my paragraphs. 1. In the great depression, many countries tried to improve their condition through the so-called beggar-thy-neighbor policy, which failed in its intention and made things worse for all. 2. Formal economic theory was developed to explain #1 above in the form of the theory of comparative advantage, showing that free international trade is optimal in the sense that world income and output will be maximized. 3. In its simplest form/model of two countries and two factors of production, the theory is associated with the names of Heckscher and Ohlin [H-O] and is based on the following assumptions. 4. Fixed supplies of labor (giving one dimension of a box diagram). 5. Fixed supply of capital (giving the second dimension of the box diagram). 6. International immobility of capital. 7. International immobility of labor. 8. Full employment of labor. 9. Full employment of capital. 10. Production functions subject to constant returns to scale, identical throughout the world, smooth, and differentiable.

Capitalism, Economic Democracy, and Ecological Destruction of Our Planet

297

11. For a wide variety of – but not all – conditions the theory based on these assumptions just stated, free trade will lead to factor price equalization among countries. 12. Using the conclusion of factor price equalization and all the underlying assumptions, the H-O theory can be generalized for a large number of factors, products and countries into the Heckscher–Ohlin–Vanek theory. 13. For a theoretical representation to hold or to be useful in understanding world economic phenomena – and to support the present day free-trade policy – it is desirable that the assumptions and conclusions used or derived would correspond to observed real conditions and observations. 14. From the point of view of observable reality, all the assumptions of the theory – except #7 are incorrect. Thus a priori the theory must be rejected. 15. And thus the liberal free trade policies of today’s world and the tenets of the World Trade Organization must be rejected and questioned. 16. The theory and its recommendations may have been at least partially valid in the context of the great depression, where the conditions of factor price equalization may have been approximated among the advanced economies involved at that time. 17. However the ‘‘signature’’ of the theory, factor price equalization, is patently untrue and unrealized in the present-day situation where wages between rich and poor countries can reach one thousand percent and beyond, in a world where for practical purposes the supply of labor in the low wage part of the world is inﬁnite (quoting the Bible, ‘‘the poor will always be with us’’). The ‘‘thousand percent’’ differential is fundamentally exogenous to the trading system, not determined by the factor price equalization theorems. 18. Under such conditions, the original theory of comparative advantage must be replaced with a much truer and better theory that I refer to as the theory of destructive trade. That theory leads to the conclusions that under free international trade and mobility of capital and immobility of labor, the capital owners will gain a lot, especially in the United States, imports of transportable goods will increase and labor employment and labor incomes in the United States will decline, especially in the import competing sectors: precisely the tendencies observed in the United States. The standard claim that free trade will be beneﬁcial to the world economy is by no means guaranteed. The only deﬁnite conclusion is that the world distribution of income, especially in the United States, will favor capital. The destructive trade theory is nothing but a simple application of the Second-Best theory where certain exogenous variables do not correspond to optimal solutions.

298

JAROSLAV VANEK

19. While I have produced the formal demonstration of a barter solution (similar to that used for the H-O theory) elsewhere, the simple version thereof can be summed up – representing the rich part of the world by the United States and the poor part by all the underdeveloped and developing poor world – the Less Developed Countries – as follows. 20. Why should anything that is transportable (TG ¼ transportable goods) in the long run be produced, under free trade and mobility of capital, with one thousand percent wage differentials, in the United States and not in the LDCs? 21. The forces of destructive trade implied by #20 are precisely what have been prevailing in the world over the recent decades. 22. The forces of reduction of production of the TGs and reduction of employment of labor was mitigated by Mr. Greenspan’s ‘‘cover-up ‘‘ by virtually zero real interest rates and overexpansion of not transportable goods, especially housing construction, Macdonald’s and consumer credit. 23. Neglecting Tinbergen’s principles of matching policy instruments and objectives, relying only on credit and interest rates, not using other policy tools, assisted by greedy users and developers of toxic debt instruments, the situation had to explode ultimately in the subprime crisis in which we are living now. 24. The crisis is far more serious than the Obama advisors are suggesting, because through the explosion we have ‘‘killed’’ the economy of the not transportable goods while still experiencing the ‘‘killing’’ of the transportable goods in the United States through the latent force of Destructive Trade. 25. It seems to me that half-cooked policies and strategies will not sufﬁce in curing the economy. Perhaps the predicted December 2009 unemployment of 8.1% reached in February 2009 can give us some insight.

GENERAL CONCLUSION The theory of comparative advantage as applied to real-world phenomena – and the practical free-trade recommendations must be rejected or at least questioned. The more realistic, even if less elegant mathematically, theory of Destructive Trade should be a better guide in deﬁning world trade solutions and leading us in the long run out of the present global crisis.

THE CASE FOR CAPITALISM: A COMMENT ON JAROSLAV VANEK’S ‘‘CAPITALISM, ECONOMIC DEMOCRACY, AND ECOLOGICAL DESTRUCTION OF OUR PLANET’’ Jed DeVaro and Adrian Stoian Jaroslav Vanek argues that the ecological degradation of the planet is more severe in a capitalist system than it would be in a fully democratic one. At the heart of the ecological preservation question is how the classic public goods problem can be solved, and we are skeptical that a fully democratic system could solve this problem any better than the capitalist system, particularly given that consumer demand for environmental protection appears to be income elastic and that the capitalist system can be expected to generate higher levels of societal wealth than the democratic system. While we disagree with Vanek’s conclusion and with many of its underlying arguments, we agree that the ecological well-being of the planet is of great importance, that it is expected that economic activity will have ecological implications, and, therefore, that it is well worth comparing the expected ecological impacts of alternative socio-economic systems.

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 299–306 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012016

299

300

JED DEVARO AND ADRIAN STOIAN

Vanek highlights the desirability of family-owned ﬁrms and small geographic distances between the locations of production and consumption. He argues that such an arrangement helps to internalize externalities such as pollution, since the members of a family-owned ﬁrm, living near the point of production, have a greater incentive to protect the environment from negative production externalities (e.g., pollution) than do shareholders, most of whom are dispersed over a wide geographic area and will not live near the externality-heavy point of production. Thus, the democratic system is less likely than the capitalist system to produce output levels exceeding the social optimum. We think that local production tends to be higher-cost, more resourceintensive production, which is unlikely to be good for the planet. Scale economies allow for more efﬁcient production using fewer resources, so in this sense capitalism might be better for the environment than a system based on localized production. Furthermore, Vanek’s argument applies to local pollutants but not to global ones. For example, because carbon emissions do not cause their harm locally, local owners would not care about such emissions any more than dispersed owners. Even in the case of local pollutants, the local inhabitants need not be owners to exert at least some inﬂuence on a ﬁrm’s decisions regarding production and the externality. A capitalist ﬁrm, after all, must hire workers to produce. Many of these workers will live near the point of production, and they have tastes for clean air and water just like the owners in a democratic system. They also choose where to live and where to work. Given fairly reasonable assumptions on information, worker mobility, and competition in the labor market, capitalist ﬁrms in a free market have an incentive to recognize the preferences of workers, since these preferences affect the ﬁrm’s costs. In equilibrium, a compensating differential will emerge in which workers living near the production point will demand higher wages. Given that economies of scale are likely to arise in pollution abatement, the capitalist ﬁrm may mitigate the negative externality to lessen the need for ongoing higher wages. One might worry that the capitalist ﬁrm might give undue attention to short term proﬁts, neglecting the long term, but with proper delineation and enforcement of property rights the capitalist ﬁrm can be expected to bear the future costs of its decisions. Vanek further argues that in a democratic system labor would enjoy higher real wages and levels of well-being than in a capitalist system. His argument suggests that the family-owned ﬁrm is insulated from competitive pressures in its effort to maximize its family’s well-being. But the familyowned ﬁrm must sell its product in a market where consumers demand

The Case for Capitalism

301

lower prices, so it is not obvious how the family-owned enterprise can indulge a taste for ecological protection any more than a capitalist ﬁrm. Furthermore, the members of a family-owned ﬁrm care not just about ecological protection but also about eating and consuming goods, so ‘‘maximizing the family’s wellbeing’’ does not mean exclusively focusing on ecological effects. The notion that capitalist societies are only about proﬁts, consumption, and materialism contradicts some basic observations. Consider the fact that countries around the world with essentially capitalist systems have developed national parks to protect habitat, and there have been many successful campaigns targeted at preserving endangered species and providing other public goods. The wealthier a society becomes, the more it can afford to allocate its resources to enhance and protect the environment. Furthermore, innovation and entrepreneurship drive technological improvements that can make transportation cheaper and ‘‘greener’’ over time. We do not see why consumer well-being would be higher in the system Vanek proposes than in a capitalist one, and we should add that any discussion of well-being must account for the prices consumers must pay for goods and services. Vanek further argues that transportation costs and the concomitant environmental degradation are higher in a capitalist system than in a democratic system. While this argument seems plausible, if we narrowly focus on transportation costs, we lose sight of some larger tradeoffs. We already mentioned efﬁciencies arising from economies of scale. In addition, capitalism facilitates innovation and entrepreneurship, which in turn spur economic growth. One must ask what the prospects are for innovation if a system based on family-oriented sole proprietorships replaces the capitalist one. Our fear is that the rate of innovation would suffer. Innovation requires an efﬁcient capital market, and an infrastructure, which naturally implies some agglomeration to achieve critical mass.1 The innovative ideas needed to advance in science and other areas of human endeavor require the synergies that are sparked when human and capital resources are concentrated. Vanek’s recommendations of decentralization and saving on transportation costs are at odds with this. We struggle to conceive of a Silicon Valley taking root in the type of system Vanek proposes. And we cannot resist feeling that it was not by accident that the Industrial Revolution was born in Britain rather than China. While we believe that the decentralization Vanek advocates may stiﬂe innovation and entrepreneurship in general, in some industrial sectors the case for decentralization might appear stronger. For example, in agriculture the use of pesticides on large acreage and genetically modiﬁed seeds may

302

JED DEVARO AND ADRIAN STOIAN

have adverse health effects, and the quality of produce deteriorates when it is transported over large distances. However, there is no guarantee that small farmers will use fewer pesticides than large farmers.2 Furthermore, transportation costs from farms to grocery stores appear to be less signiﬁcant than energy costs associated with storage and food preparation, which are in turn likely higher in a decentralized system due to inefﬁciencies in household consumption of food; there is also a cost advantage to largescale agricultural production. As a consequence, the decentralized approach is likely to imply higher prices and, due to less transportation, fewer options, both of which will hurt consumers.3 At the same time, as consumers become wealthier, their demand for products like local, organic fruits and vegetables and free-range, grass-fed beef increases, that is, demand for these goods (as for pollution abatement) is income elastic. A shift toward a more decentralized approach to agriculture could occur in the context of a capitalist system, as evidenced by a growing number of farmers markets and community-supported agricultural and organic grocery stores and restaurants that feature local produce. Wealth in capitalist societies has increased to the point that, for example, consumer attention increasingly turns to the conditions under which chickens and cows live before being slaughtered for food. These are high-income-elasticity issues that would not have occupied the minds of the typical 18th-century consumer, and our feeling is that issues such as California’s 2008 ‘‘Proposition 2’’ resonate with voters precisely because of the capitalist system and the wealth it creates, rather than despite it.4 Shifts from centralized to decentralized production, motivated on ecological grounds, must be driven by consumer demand, and the higher average incomes in a capitalist system can be expected to facilitate that. Vanek emphasizes that the location of production matters for the global ecological footprint. We argue that from a global perspective and over a longer term, it should not matter whether the pollution emanates from China or from the United States. In the short run global pollution might increase when production moves from a developed economy like the United States to a developing one like China, in part because the poor in developing economies have weaker preferences for environmental quality than do the rich in developed countries. However, given that demand for environmental quality is income elastic, over time the increasing wealth in developing countries can be expected to lead to increased demand for environmental protection. While this may take a long time, in the shorter run there is another mitigating factor.5 Those (richer) consumers in developed countries will tend to prefer environmentally friendly production practices. So the innovative capitalist ﬁrm can be expected to devise ever more efﬁcient ways

The Case for Capitalism

303

to produce in environmentally friendly fashion and to appropriately inform consumers, reaping a markup through product differentiation. The consumer information assumptions required for this argument to hold are increasingly reasonable in the age of the Internet. Vanek also comments on the distributional consequences of trade for domestic labor, although its relevance to his main theme of ecological degradation should be clariﬁed. Obviously trade has distributional consequences, and in the United States in recent decades the wages of low-skilled labor in some sectors have been pressured by trade. At the same time, trade intensiﬁes demand for U.S. exports which creates new demand for domestic labor.6 We also note that the wages paid to domestic labor are only one side of the coin; on the other side are the prices these workers face for the products they consume. All consumers welcome the lower product prices facilitated by trade, including those family members who would work in Vanek’s family-owned ﬁrms (and who care about eating and consuming as well as about ecological degradation). Vanek’s arguments create the impression that the democratic ﬁrm is somehow insulated from competition. But how would such ﬁrms compete with cheaper imports? What is the likelihood that they would have fared better than their capitalist counterparts once Deng Xiao Ping started the process of reform in China?7 From an American standpoint, it seems fortunate that American companies had the scale and the backing of capital markets to quickly deploy resources to China for investment and to seize the new opportunities in that part of the world. Finally, if we take a global perspective when looking at the wages of low-skilled labor, rather than focusing narrowly on the U.S. experience, we must balance the pressure on the real wage of U.S. laborers against the tens of millions who escaped poverty in developing countries including China. This is cause for celebration for opponents of global wage inequality. Vanek also argues that proﬁt maximization is not desirable because labor costs and use tend to be minimized. In 1900, half of the labor force in the United States worked in agriculture, versus less than 4 percent today. Labor productivity in agriculture has increased dramatically during the last century. But imagine telling people in 1900 that in less than a century there would be less than 4 percent of the workforce in agriculture. This would likely have incited fears that the U.S. economy would be paralyzed with an unemployment rate approaching 50 percent, due to minimization of labor costs. But such fears ignore the role that innovation and technological improvements in a capitalist system play in expanding the set of opportunities. We believe that Vanek’s proposed system is less likely than capitalism to generate these beneﬁts and that a better way forward is to foster an environment of creativity

304

JED DEVARO AND ADRIAN STOIAN

and entrepreneurship, while supporting people’s acquisition of the skills needed to succeed in a highly technological, innovative and globalized 21stcentury economy. We see no evidence suggesting that product quality will be worse in a capitalist system than in a democratic one. The capitalist system produces cars such as BMW that many in the world dream of owning, yet the Trabant (the East German car of choice until its production was discontinued after the crumbling of the Berlin Wall) was, it is fair to say, somewhat less appealing than a BMW. Product quality is driven by consumer preferences; if there is a demand for quality in the capitalist system, the capitalist ﬁrm will provide it. Furthermore, if what consumers prefer is a lower-quality, cheaper product, then how is anyone better off if what is provided is a higher-quality, more expensive product? The demand for high-quality products, like the earlier demands we have noted, is income elastic. We would expect that the wealth generated by a capitalist system will lead to increasing demands for product quality, but quality must be driven by consumer demand y not the other way around. Vanek argues that the welfare-maximizing democratic ﬁrm can control its working hours to a greater extent than the proﬁt-maximizing capitalist ﬁrm, thereby producing less, using fewer resources, and causing less environmental degradation. He further argues that workers in the democratic system can consume more leisure, since ‘‘labor incomes per hour can be higher than the capitalist wage.’’ These arguments neglect several points. First, workers with varying tastes for work versus leisure and for workplace conditions can select among employers, so that competing proﬁt-maximizing employers must recognize the preferences of workers. Employers choosing to ‘‘set’’ hours without regard to worker preferences will meet an unhappy fate in a competitive environment. Second, Vanek focuses on wages but ignores product prices. His language suggests that market competition only drives down wages and not product prices. Consumers care not about what wage they are paid but rather about what that wage can buy for them. Finally, the notion that workers in a democratic system can be expected to ‘‘consume more leisure’’ might well be true, if we interpret leisure as involuntary unemployment. By stiﬂing innovation, entrepreneurship, and the resulting growth that creates jobs, the democratic system can be expected to have a depressive effect on employment.8 To conclude, we are skeptical that a fully democratic system would be better than a capitalist system at fostering the ecological welfare of the planet and enhancing human welfare. The fully democratic system with its family-owned ﬁrms and restricted trade seems unlikely to encourage global

305

The Case for Capitalism

cooperation on issues like climate change. We believe a capitalist system is more likely than the alternative to generate the greatest amount of societal wealth and that as a society’s wealth increases so does the strength of its preferences favoring ecological protection. To be clear, we are not arguing that economic activity has no negative consequences for the environment or that these consequences are not serious and cause for potential concern and possible government intervention.9 We are merely arguing that Vanek’s alternative, to us, is not obviously better on the ecological front, and is clearly worse on other fronts. That being said, we end where we began, by reiterating that the broad point Vanek raises – that is, that alternative economic systems may have differential ecological effects on the planet – is important and worthy of dispassionate and reasoned debate.

NOTES 1. This agglomeration often results in the formation of cities that are more energy efﬁcient than decentralized spatial organizations. 2. For example large farmers can afford to invest in sophisticated technology like satellite imaging that facilitates a more efﬁcient use of inputs, including pesticides and – importantly in many regions – water. 3. As an example illustrating limited choice, consider that the cost to Oklahoma of producing its own Pinot noir would be prohibitively high. 4. Proposition 2 (Standards for Conﬁning Farm Animals) was on the November 4, 2008, ballot in California and passed with 63.5 percent of the vote. The proposition created a new state law – which goes into full effect on January 1, 2015 – prohibiting the conﬁnement of farm animals in a manner that does not allow them to turn around freely, lie down, stand up, and fully extend their limbs. The proposition marked the ﬁrst time that California voters were called upon to vote on such a question. 5. The process is likely to take a longer time in countries where institutions are weak, and the relationship between income and environmental protection might be non-monotonic. Furthermore, even looking within countries, there could be considerable variation across different types of pollution in the speed with which environmental protections are implemented. In the long run, however, the prospects for global environmental protection should improve as countries become wealthier. 6. This is true even for American products produced abroad. In recent years, some American companies have increased sales of their products in China. Even though production of these goods is often ﬁnalized in China, the growing sales in China create demand for American labor since a substantial share of the total labor costs is associated with countries other than China, including the United States. 7. Increases in competition naturally lead some to argue for protectionism. However, our sense is that autarkic economies have fared worse than open ones, ceteris paribus. By any measure, China today is more open than China few decades

306

JED DEVARO AND ADRIAN STOIAN

or centuries ago, and the consequences of this for its rate of economic growth are clear. 8. Also on the subject of leisure, we note that labor-saving home appliances such as washing machines and microwaves allow consumers to spend more time on recreational activities. These products too are the result of innovation and entrepreneurship, which again are fostered by capitalism. 9. Such intervention would be of the usual form – for example, Pigovian taxes and subsidies, tradable emissions permits, or World Trade Organization regulations concerning production practices – and might have a role in both the capitalist and fully democratic systems.

REJOINDER Jaroslav Vanek I welcome any dialogue on the present subject because it clariﬁes and makes us reﬂect on such a signiﬁcant issue of our times. Especially I would like to stress the subject of the appendix on destructive trade. We live in a world impregnated with the notion of free trade optimality, but in many contexts the notion is incorrect, and such mistakes can be serious for the advanced economies including the United States. For the theoretical mind of economists, let me stress that the theory of optimal comparative advantage includes the conditions of full employment and factor price equalization [or near-equalization], but with wage differentials of the order of one thousand percent the optimality is most unlikely. The unemployment and the ﬁnancial crisis of the great recession are the most dramatic and convincing veriﬁcation – as well as substantiation of my arguments concerning global ecology.

Advances in the Economic Analysis of Participatory and Labor-Managed Firms, Volume 12, 307 Copyright r 2011 by Emerald Group Publishing Limited All rights of reproduction in any form reserved ISSN: 0885-3339/doi:10.1108/S0885-3339(2011)0000012017

307